Ulysses by Numbers

Why is Joyce’s Ulysses as long as it is? Too often we forget that the size of this book was not somehow written in the stars. It began, modestly enough, as a sequel to the 85,077-word A Portrait of the Artist as a Young Man, but in the course of seven years, it was transformed from a novel with modest ambitions into that morphological nightmare made up of 264,448 graphic units (or language tokens); 39,801 different entries; and 25,541 lines of text.¹ What happened? How did it go from being a book to that “fat book” as Vladimir Nabokov once called it?² And, in the process of trying to answer a question about size, we might want to think about that other unsolved mystery: What formal and material factors may have actually determined the length of the individual episodes?

That’s the subject of this chapter: a quantitative examination of Ulysses that must, by necessity, grapple with a number of genetic issues involving its evolution through a series of early, middle, and late stages that together chart the transformation from a more novelistic style mixing interior monologue and third-person narration to the experimental theatrics characterized by a series of narrators with a lot of personality. So, as a first step, I need to clarify what version of Ulysses I’m talking about. We already know a great deal about the book as it grew from the notes to the notebooks, from handwritten drafts to typescripts, from faircopy manuscripts to proofs. What we don’t know much about is the serialization: the thirteen and one-quarter episodes that appeared in the Little Review between March 1918 and December 1920 in twenty-three different installments.3 For decades, critics have largely dismissed the relevance of these versions because of the inaccuracies and excisions, and it’s only recently that a publisher even thought it worthwhile to reprint them in a single volume.⁴ The regular deadlines, they concede, may have helped Joyce get the episodes written (even if he kept notes to add later on), but the transmission of the text from typists in Trieste and Zurich to Ezra Pound, then to the relay in London (who corrected and sometimes censored them), and on to the compositor in New York (a non-English-speaking Serbian) ended up generating a mutilated version of the novel not worth its weight in pulp.⁵

That, of course, is the problem. Those critics interested only in the accuracy of the text—the ones in search of an ever-elusive authoritative version—have been missing out on what’s important about the serial Ulysses. Indeed, the Little Review may have been responsible, in part, for textual corruption as the novel was appearing, but it remains relevant precisely because it was the medium that first gave this novel a home, and, more important, it shaped the very scale of the episodes and determined the tempo as they came out on a monthly, and later bimonthly and trimonthly, basis.⁶ Joyce had been taking notes and drafting versions of various episodes at least three years before the serialization began, but once the clock was ticking and installments were due, he kicked into a regular routine: seven episodes appeared between March and December 1918. That’s more than half the total number of episodes published in the Little Review before the process was repeatedly delayed by paper costs, meddling censors, and timid printers and then eventually ended by legal battles.

Before moving ahead, it’s worth asking what an episode is in the first place.7 The term is common parlance in Joyce circles, one of those words used to show you’re in the know. I suspect that the easy adoption had a lot to do with the fact that each installment in the Little Review was defined as an episode (without the Homeric titles by which Joyce referred to them) or, in some cases, defined as part of an episode and followed by a roman numeral. The word episode, however, does not carry the same meaning as section or chapter.⁸ These are bibliographical terms regularly used to identify the internal structure within a book. And it could be any book: the Bible, the plays of William Shakespeare, the essays of Samuel Johnson, or the poems of John Milton.

But episode as a category for the structural division within a book works differently. The Oxford English Dictionary defines it more generally as a digression within a story or poem.⁹ Episodes, in other words, are a detachable part of a narrative that can be reintegrated into a larger whole. The serialization of novels, which involved separate installments identified as numbers, ended up benefitting from episodic organization, in part because they could provide places in the plot for installments to break. Episodes within a narrative became a way of thinking about how the novel, which could be long and unwieldy and difficult to remember, was organized. In other words, if readers wanted to identify something they had read in the course of a year or two, the episode in which x, y, or z happened was the best indicator, because it did not rely on a numbering system that very few readers could keep track of.

Episode, then, defines a structural element within the novel that gets associated with a serial mode of novelistic production: episode is not chapter, and, for that reason, it does not depend on the organizational logic of the book for its meaning. The episodic appearance of a novel within a newspaper or little magazine was made possible by serial modes of production with built-in temporal interruptions. Every week, fortnight, or month could bring the continuation of a story, but if you think about it, this passage of time in between would invariably affect how this story was written and received: readers had to keep track of the plots and the characters, which is not an easy task when the serialization process is attenuated. And behind it all, the episodic experience would influence how readers might even imagine the novel itself: Was it the assemblage of fragments released over time in a newspaper or periodical? Do you even need the book for the novel to exist? More to the point, what is the difference between the version housed between two blue covers and the one that appeared serially in the Little Review? In asking these questions I’m not in search of an authoritative text; I’m interested in the medium through which that text was transmitted. Thinking about the medium of the little magazine raises some provocative questions about what Ulysses is and is not, one of them involving length: Why are the episodes in the Little Review as long as they are (which gets us to the bigger question I posed at the beginning)?

Measuring the length of the serial Ulysses, simple as it sounds, is not such a straightforward exercise. In the process of trying to figure this out, I considered three possibilities: the pages of typescript (sent by Joyce to Ezra Pound and distributed to Margaret Anderson, editor and publisher of the Little Review, in the United States), the pages of the faircopy manuscript (drafted by Joyce between 1918 and 1920 and sold to the lawyer and collector John Quinn), and the printed pages of the episodes in the Little Review (averaging sixty-four per issue). The typescript isn’t reliable because Joyce depended on too many different typists with different habits over the years, too many different typewriters with different spacing, too many different kinds of paper with different line counts, and, on top of it all, there are too many missing pages, episodes, and/or mixed copies. The faircopy manuscript (also known as the Rosenbach manuscript), which Joyce produced by hand (sometimes relying on the generosity of others), does not always correspond directly with the typescript that was used to set the pages of the Little Review, and Gabler has suggested that a now lost manuscript was used for many of them.10 Counting the pages of the Little Review might seem like the most viable option, but the page layouts, formatting, line length, and line numbers change too frequently over the years, making it impossible to establish a reliable standard for measurement.¹¹

After a lot of searching, I finally realized that I was getting it all wrong. Inspired at first by Gérard Genette’s analysis of Proust’s À la recherche, I was looking for pages to measure when it was really words I was after.12 Kenner was on to the idea of word counts decades ago when he made the observation in passing that I used as my epigraph. Because words are blocks “delimited by spaces,” we can indeed count them. Kenner’s own calculations with word blocks led him to a number of conclusions, including the one, mentioned earlier, that the twenty-two words in the first sentence of the novel correspond both with Homer’s hexameter (two lines really) and the number of books in Leopold Bloom’s library (excluding the one owned by the Capel Street Library that he never returned).¹³ But in addition to these more eccentric local correspondences, Kenner also had his eye on the big picture, pointing out that if Joyce had stopped with episode 10, Ulysses would have ended up as “a moderately orthodox novel of fewer than 100,000 words, its ten chapters each of fairly normal length.”¹⁴ Kenner’s word count was off by a quarter here (74,720 words is the real total), but the point he wanted to make is still valid: the bulk of Ulysses was something that happened gradually, month to month (with occasional lags in between), episode by episode, installment by installment. Joyce was calibrating and recalibrating the size of his novel as he went, and it wasn’t just the book that was getting longer, it was the episodes themselves as the book was being written.

After receiving the electronic files generated from Gabler’s critical and synoptic edition of Ulysses, Kenner was reminded “how deeply our experience of Ulysses integrates with a printed text.”¹⁵ Counting and reading are possible precisely because there are word terminations still waiting to be measured if we ever want to get into the novel’s numerical unconscious. And the counting must go on.¹⁶ Of course, I’m not the only one to embark on such an extensive numerical adventure. Long before Kenner brought out his calculator, there was Miles Hanley’s legendary word index, first published in 1937 and intended (according to Martin Joos) for the student of “association psychology” tracing image patterns, the Joycean looking for “verbal echoes,” and the educators and linguistic scientists decoding the “structure of vocabulary.”¹⁷ Almost a half century later, when Gabler and his team assembled the corrected text, an alphabetical index was generated from the “output files.” Keen to distinguish his updated “handlist” from the convenience of a concordance or dictionary, Gabler believed that the user might “gain both excitement and insight from exploring on his own the paths of the philologist that often enough lead to discoveries of some moment in the language universe of Ulysses.”¹⁸ If the word was going to be the philological key to this “language universe,” the handlist was there to help pick the locks.

Delivering one of the plenary lectures on the occasion of Gabler’s edition of Ulysses, even Jacques Derrida got into the spirit of things by tracing the frequency of a single word throughout the entire novel. Strangely enough, he seems to have ignored the handlist altogether, choosing instead to do the counting himself, one yes at a time. “I myself computed, a pencil in my hand,” he told the audience, “the mechanical figure of yeses legible in the original gives more than 222 in all, of which, more than a quarter, at least 79, are in Molly’s so-called monologue(!).”19 If the image of a pencil-wielding Derrida counting yeses one by one is not entertaining enough, it’s worth pointing out that he still ended up getting the final number wrong: there are actually 359 “yeses” in Ulysses (91 of them in Molly’s monologue), and that doesn’t even take into account the yups, yeas, sis, ouis, and ays.²⁰ Which, I suspect, would have been fine with Derrida, since he wanted to demonstrate above all that computer-generated lists, though convenient, are incapable of dealing with the slipperiness of signification for any given word.

Over the decades, the focus on counting has been placed on the book Ulysses, and in each case there has been an emphasis on the philological value of tools like word lists, handlists, and indexes. That is not my goal here. Counting words in the serial Ulysses is one way we can understand how it was happening in real time over the course of two years. Ulysses demands to be read this way. Eliot may have suggested early on that a “mythical method” is what held the various parts of this novel together, but there is, I would argue, an architecture that was necessarily generated by the unit of the word itself. Joyce wasn’t just writing month after month in order to meet the deadlines, he was measuring as he wrote, and, even if he would later revise each episode for Darantière in 1921, the early structure for a majority of them was firmly in place. Counting is a way of containing, and the French word for storytelling, raconter, is an appropriate reminder here of just how much the two are entwined. To narrate is to count time passing, and, when words are actually inscribed on a page (instead of being uttered by a storyteller), that passage of time, that measurement, is visible to the naked eye. We may like to fashion Joyce as the master raconteur, but when doing so we need to remember that during the serial stage in the evolution of Ulysses, he was also a mechanical counter, someone eager to keep track of how many words he was using as the plot itself was unfolding one episode at a time.

So, a limit of “6,000 words” per installment: How long is that? If the average word count per page in the Little Review is roughly 350, then it would be somewhere around seventeen pages (and, in fact, the first installment was nineteen pages). Pound’s limit of “up to 6,000 words” was, for Joyce, an approximate number, “about 6,000,” that could be stretched a bit if necessary. When the first episode arrived at Pound’s flat in London, these seventeen densely typed pages were made up of 7,001 graphic units, and in the few weeks and then months that followed Pound received four shorter “bundles” of typescript averaging 4,800 words each.24 Before continuing, I should mention that these numbers were generated from the same files used by TUSTEP (an acronym for Tübingen System of Text Processing Programs) that was specially designed for the critical and synoptic edition of Ulysses. Gabler generously passed his master files along to me, and, in preparing what follows, I ran them again through another word-counter program and double-checked them against those of a computational analyst: we both arrived at slightly lower numbers than Gabler’s. It turns out that the dashes used to identify bits of dialogue in the Little Review files are largely to blame: they were mistakenly counted as tokens along with some of the symbols introduced during the reformatting process to distinguish foreign words.²⁵ From now on I will use the terms word/word count, though I’m referring to tokens, or what you might also call graphic units. The computer counts the spaces between blocks of text to determine a word, so it is the spaces that signal where each unit begins and ends. Initials like J. C., K. M. R. I. A., E. L. Y’s, and I.O.U.’s will make up a single unit and so too will the hyphenated words, portmanteaus, abbreviations, titles, units of measurement, numbers, abbreviations (i.e., c/o), and stand-alone letters.

I want to concentrate now on the length of the first six episodes (figure 2.1). Together they make up Stephen Dedalus’s “Telemachia” (episodes 1–3) and the first three episodes of Leopold Bloom’s “Odyssey” (episodes 4–6) The episodes in each sequence begin at eight, ten, and eleven in the morning, respectively, and, as David Hayman was the first to notice, they mirror each other structurally and thematically. But there’s more going on here: the elements of this pair of three episodes each are actually the same length. Stephen gets, on average, 5,578 words per episode, Bloom gets 5,940. Averaging the episodes in this early stage of Ulysses is important because it reminds us that word count does not have to be restricted only to isolated episodes. Therefore, even if the word count of episode 1 is higher than that of the two episodes that follow (which are framed by the higher episode 6 on the other end), the combined average is still just below the limit: and with the three episodes of Bloom, the same thing is true. There is episode 6 on the high end preceded by two episodes below 5,000: together that’s “up to” or “about” 6,000 words, the number that Joyce and Pound had originally agreed on.²⁶

But why 6,000? Where does that number even come from? Dante used between 115 and 157 lines for the Inferno (eleven different canto lengths, Joan Ferrante argues, so the chaos of Hell could be properly represented in the poem’s form); Shakespeare, writing plays too long to be performed in their entirety, hovered somewhere around 3,000 lines (when his contemporaries stayed closer to 2,430), and Dickens stumbled upon a twenty-part, nineteen-month format, in thirty-two-page installments, that he would use for eight of his major novels. Six thousand words, however, works differently as a unit of measurement: it was identified at first as the necessary length for what was still an undetermined number of episodes.²⁷ Not only was there no precedent at this moment when 6,000 was decided upon, but there was also no end in sight. Six thousand words multiplied by what? Where, finally, would the accumulation of this number lead? Pound had recommended it at first as an upper limit, and, when he did so, he was asking Joyce for a short story, not the serialized installment from a novel. Once they agreed to bring out Ulysses in the Little Review (with the possibility that it would appear simultaneously in the Egoist), Joyce still latched on to that number, deciding that 6,000 words per episode sounded about right. Was that the average length for a serial installment in little magazines at the time? Was it an average left over by the Victorians?

None of the above: there was no average word count for serialized novels in little magazines because so many of them didn’t actually pay in the first place, especially not by the word. And if there were modernists looking to the Victorians for serial guidance, I doubt that length would have been the primary interest. I want to suggest another possible explanation that might, at first, seem preposterous. A total of about 6,000 words is the average episode length in the 1879 Butcher and Lang prose translation of Homer’s Odyssey, one that Joyce, we know from his friend Frank Budgen, relied on while living and writing in Zurich (figure 2.2).²⁸ And even if, like Kenner, you disagree with the assertion that this was one of the Homeric sources Joyce relied on, preferring instead the 1900 Samuel Butler translation that was later recommended by Stanislaus Joyce, then you still get a close word count. It should be added, though, that because Butler is more concerned with the substance of his translation than an ornamental style, the word count in his translation is consistently lower.²⁹

I’m fully aware that the Odyssey was not serialized, but it was episodic, and the tradition of oral recitation of different books at different times by different poets was, in itself, an early mode of serialization. Henry Fielding made that same point in Joseph Andrews, when he joked that Homer “was the first inventor of the art which hath so long lain dormant, of publishing by numbers!”30 Joyce, I suspect, would have been tickled by the idea: it would have been one more bond between bards. But considering the intense scrutiny Joyce paid to all kinds of esoteric symbols, strange parallels, and tiny details in the Odyssey, it’s not ridiculous to suspect, even in the early stages of Joyce’s work, that length would have come up along the way as well. The Butcher and Lang and Butler translations do not include word counts on the page, but both of them have line counts for each of the twenty-four books that would have made these calculations very easy: roughly 12 words per line multiplied by 444 lines, for instance, make up book 1: that’s 5,328 estimated words.³¹ The rest of the books in the Odyssey vary between 5,000 and 6,500 words each, with a single outlier in book 4 (9,491). Added together, the average is 5,645 words per book, which is 68 words more than the average for the episodes in the “Telemachia” and 295 words fewer than the average for the first three episodes of Bloom’s “Odyssey.”

Had the episodes in Ulysses stayed within that 6,000-word limit, the riddle about length might have been solved. Joyce, we might have been able to suggest, took a Homeric number and ran with it all the way back to Ithaca (episode 17). That, of course, is not how all of this plays out. The length spikes with episode 6 and the following two episodes, when the number of words surges from an average of 5,233 to 8,813 (figure 2.3). At this point in the composition of the novel, Joyce doesn’t just exceed the projected word count he and Pound had settled on, he nearly doubles it. At 8,813 is almost twice the average of episodes 2, 3, 4, and 5 (4,791 words). So what’s happened here? Why this sudden spike in word count, which effectively, and from that moment forward, changed the terms of Joyce and Pound’s serial agreement?

The short answer: Dublin. With episode 6 the space of Ulysses suddenly expands, and what Moretti has said elsewhere about the inextricability of adventure plots and novel length applies here as well: an increase in word count is directly related to the widening of space in this novel.32 It’s a shift that needs words for the plot to move forward, sideways, backwards, and around. Before episode 6, Bloom’s movements are severely restricted: a short walk to the butcher shop in episode 4 followed by a discrete stroll down a few streets in episode 5, but once he gets into the carriage headed to Paddy Dignam’s funeral, Dublin suddenly dilates. It is the first extended view of the city (again in episodes 7 and 10), one that officially marks Bloom’s headlong leap into the urban fabric. And Bloom is not just in Dublin here as the carriage winds its way through the streets: he is beset on all sides by Dubliners. The restricted social interaction of the earlier episodes, which could easily be condensed into 6,000 words or fewer when he was mostly in his head, gives way to the formation of an expanding network of relationships and new spaces that simply can’t fit. And Paddy Dignam’s funeral is just the beginning. From that point forward, the cast of characters increases considerably, and the number of different locations proliferates: the newspaper office, Davy Byrne’s pub, the library, the Ormond Hotel, the cabman’s shelter, all the way back to 7 Eccles Street.

Doubling and Dublin, then: a spike in word count, I’m suggesting here, was generated by the spatial expansion of the plot (figure 2.4). With the publication of episode 6 in the Little Review it’s as if the artificial constraints first recommended by Pound and seconded by Joyce could no longer contain the scope of the novel as the adventures were developing. And it’s worth reiterating here that the serial Ulysses was a live event: Joyce may have had a rough plan for the number of episodes and the order in which they were to appear, but once the serialization process kicked in, he was still trying to figure out how to piece it all together month by month both within and between episodes. Even with a general plan, and the notebooks he kept with words, phrases, and fragments of paragraphs, these episodes were taking shape as they were being written with deadlines in mind. By July 1918, when Joyce sends Pound the typescript for episode 6, about 9,000 words was the new limit, and it lasted for two more episodes (episodes 7 and 8).

And then another jump: from episode 9 through episode 11, the average word count rises from 8,813 to 11,179. As was true the first time the word count jumped, this spike is not followed by a sudden jump back to the earlier limit: there’s another phase of leveling out for two episodes (and then another jump). It’s worth mentioning that in both instances, these moments of leveling out contain three episodes each. It’s impossible to say with any certainty why three is the magic number (though I’m sure Kenner would have had a compelling theory), but it seems as if leveling out was a way to normalize the length increase during the serialization process, as if Joyce exceeded the limit originally established and then attempted to justify what he’d done by repeating it twice before doing it again. In the end, that repetition makes the length increase appear to be more of a calculated necessity than a serial aberration.33

As evidenced in figure 2.3, episode 4 came in at 4,923 words: six episodes later episode 10 reaches 11,212. But just consider for a moment what it would mean if these numbers were somehow reversed: Could the serial version of episode 10 have been contained in just under 5,000 words? Considering the fact that there are these dramatic increases in series of three episodes, it’s worth thinking about. Joyce justified the formal experimentation of his novel on the grounds that each episode “conditioned its own technique,” and, looking at the distribution of words on the chart, it’s possible to make a similar claim about numbers: each episode conditioned its own word count. Joyce may have begun Ulysses with some general idea about length, but in the course of writing month-to-month, episode-to-episode, he clearly began to change his mind. The data is compelling for what it tells us about Ulysses as the novel was appearing in serial form: some things about proportion, scope, and scale could be anticipated, but word count was not one of them, and, as a result, the rules were bound to be broken again and again.

That, in effect, is the point I want to emphasize. The logic of length is something every writer has to confront. In this case, you could say that there’s an internal rhythm to the serial Ulysses, but with the production of each installment, Joyce was not just adjusting the size of individual episodes; he was changing the scale of the entire novel. Episode 4, then, to go back to the question just raised, couldn’t be the length of episode 10, and vice versa, but that’s not because Bloom’s kitchen is too small or Dublin and its cast of characters too large. Rather, it’s because they both belong to a specific moment in the serial composition of the novel, each one assuming a particular length because of the episodes that precede it, and each one will subsequently help to determine what happens as the process continues (working within the limit of 6,000 words only to break out of it).

Unlike the first spike in word count, however, this second one had serious material consequences: from that moment forward, individual episodes of Ulysses had to be broken up into two (and later three and four) different installments. And there was an obvious reason: any single episode published in its entirety would have taken up half the space of the little magazine, which averaged somewhere around 23,000 words per issue. Joyce might not have minded, but it’s likely that the other contributors and readers would have taken offense. Looking at the numbers, though, you begin to wonder if that might have been the point: not that Joyce was staging a silent takeover of the Little Review but that he finally realized a single episode could be divided between two different issues. After all, this second spike includes another sequence of three episodes that could have been broken down into two installments of between 5,000 and 6,000 words, and the number never dips below 11,000 again.34

Is this just a coincidence? Not likely, given the size and consistency of the increase. Joyce, these numbers reveal, wasn’t just making his episodes a little longer than before: he was making them big enough to appear as two separate installments of just under 6,000 words. There’s no evidence I’ve found so far to indicate that Joyce explicitly recommended this course of action, and, in their correspondence (at least the portion that has survived), Pound never mentions that he has a plan of this sort up his sleeve, though I suspect that when sending the typescript on to Anderson, he would have made some indication where each of the longer episodes could be divided.³⁵ Though we will probably never know who, finally, was behind this two-issue solution, it ended up revolutionizing the form of Ulysses: once an episode could run beyond the frame of a single issue, who was to say how long it could or should get? It was this act of working within and then moving beyond the material constraints of a little magazine like the Little Review that allowed, even enabled, the radical formal experimentation that followed.

I’ve been talking about the effect of this spike on episode length, but why does this second one happen where it does? And what can it mean to the way we think about the episodes that followed (12, 13, 14)? When I first began working through this part of the argument, I hit a snag. The quantitative data was there all right, but I was having a difficult time trying to figure out what, if anything, it could tell us about a qualitative difference in the novel. This is where the genetic component of my argument arrived in full force. I was particularly drawn to an observation made by Groden when he was first piecing together the genetic history of Ulysses in the mid-1970s using the available typescripts, manuscripts, and notebooks. Reading through what was available at the time, Groden arrived at the conclusion that the end of episode 9 represented a break with the “initial style,” one that Joyce had identified more closely with the technique of interior monologue. As evidence, he used the note that Joyce inscribed at the end of this episode when preparing the faircopy manuscript for John Quinn on New Year’s Eve, 1918, which reads: “End of Part I.” Critics after Groden have tended to agree that since this end does not correspond with the divisions of the three-part structure of the novel, it must therefore signify the self-conscious departure from the style that was being used up until this point.36

After episode 9, Joyce wasn’t just moving beyond the “initial style,” he was breaking out into the formal experimentation that would go hand in hand with the forceful return of an omniscient narrator who heckles the characters, remembers phrases and events from earlier episodes, and, before long, hijacks the story completely before Molly reasserts control. Groden suspects that this gradual discovery of the need for an omniscient narrator, which forever changed the direction of the novel, was quite unexpected for Joyce, “a result of his episode-by-episode progress on his work, progress that never extended far beyond the immediate episode at hand.”³⁷ Indeed, there are moments of interior monologue after episode 9 that come up in episodes 10, 11, and 13, but in each instance, they are made to accommodate the presence of someone or something who refuses to let any single character cogitate for too long.

Building on Groden’s observation, I want to consider in greater detail how word count might fit in. As Joyce moved away from interior monologue and back again toward omniscient narration, the episodes got even longer, and part of the trick would involve figuring out who was wordier: Stephen, Bloom, or these new unnamed narrators. Wyndham Lewis, for one, was convinced that Bloom was the culprit, an “abnormally wordy” character who just didn’t know when to shut up. This wordiness, he goes on, ends up making all of his monologues about as convincing “as a Hamlet soliloquy.”38 Rising to Joyce’s defense, Stuart Gilbert argued that Lewis was missing the point. There are, he explains, some people who believe that “without language there can be no thought,” adding further that even if we don’t think in words, “we certainly must write in words.”³⁹ What’s particularly interesting about this exchange is the way the question of wordiness gets redirected so quickly into the idea that words and thoughts are inextricable from each other: without words, the mind is a blank slate. Gilbert evades Lewis’s whole question by emphasizing that length and economy of language are necessary whenever thinking is actually in progress. Joyce, like any other novelist using interior monologue, needed to enter into the minds of his characters but was forced to use words to document what he found there. Even though he knew that brushstrokes and instantaneous impression were not an option, Lewis believed that some of their effects were still reproducible in other media.

I want to take this criticism of Bloom’s abnormal wordiness seriously. When, after all, does the gradual accumulation of words in a character’s mind make him or her seem longwinded? And is Bloom really any wordier than Stephen or the cheeky narrators that trail the two of them? Any attempt to try to answer these questions requires parsing each of the episodes in order to distinguish, as much as possible, between moments of interior monologue and omniscient third-person narration. In order to do so, I’ve isolated and counted each of them separately, including the moments of dialogue in the omniscient category, which the narrator would be privy to, along with the sentences that are part of what Kenner calls the “novelistic housekeeping,” all “the cames and wents, saids and askeds, stoods and sats, without which nothing could get done at all.”⁴⁰

And if we are really just trying to figure out who’s the wordiest Dubliner of them all (at least before the blustery episode 12 arrives), then Lewis is right: Bloom wins. Even Stephen’s wordiest episode (episode 3) is still two thousand words shy of Bloom’s (episode 8). But if we want to document the stylistic shift that Groden first identified, it’s necessary to think about the ratio between interior monologue and omniscient narration in all of the serialized episodes. Out of the thirteen that appeared in toto in the Little Review, four belong to Stephen (episodes 1, 2, 3, and 9) and seven belong to Bloom (episodes 4, 5, 6, 8, 11, 13, and 12). I’ve excluded episode 10 from the final tally because of its synoptic view, which effectively integrates Stephen and Bloom into the much larger social and urban fabric of Dublin.

Looking at the ratios, a few patterns emerge (figure 2.5). For one thing, Bloom tends to share space equally with the narrator, whereas Stephen gets eclipsed. Even more interesting is the fact that in the later episodes (11 and 13), the narrator, who is beginning to muscle in, continues to give Bloom some time with his thoughts. In fact, the number of Bloom’s words in the later episodes is roughly double what it was in earlier episodes (e.g., 4, 5 and 6), and the same is true for the narrator, who had a similar average before Bloom’s increase. By way of a provisional conclusion, these numbers suggest that the takeover from episode 9 forward was gradual. Joyce may have begun to make arrangements for the return of authorial omniscience, but he was still thinking in terms of the same ratio. Each of the three episodes is around 11,000 words, which, as I have pointed out, was double the original limit, but shows some sense on Joyce’s part that the proportion itself would not be disrupted so long as an episode could appear in two or more installments. But once the narrator no longer has to make room for the musings of Bloom or Stephen, the episodes don’t get shorter. All of this newly discovered space for more words gets used up and there is even an increase, which, in the end, will make it possible for the stylistic parodies to go on longer, in episodes 12 and 14 especially.

Take a look at what happens when the word count spikes for the third and final time with episode 12, a 16,510-word episode appearing in four different installments of 6,532; 4,319; 3,500; and 2,159 words, respectively. Episode 13 comes in at 14,231 words and appears in three installments of 3,099; 4,552; and 6,580 (figure 2.6). With episode 14, the number jumps back up to 19,214, though only one portion is published before the serialization stops for good. It’s clear here from the word counts of these three episodes that the possibility for separate installments radically changed the way Joyce was continuing to think about the length of his episodes. By the time he began drafting episode 12, he was envisioning a much fatter book with even more audacious formal experiments. How else can these proportions be explained? To me they’re a sign that Joyce may have known where the ports of call were in the novel he was writing, but he wasn’t always sure how long each of the adventures should last: it took the pressure and regularity of serialization to figure that out. And by the time he arrived at episode 12, the same one that marked the “end of the initial style definitively,” he had a pretty clear idea of how the technology of the little magazine could be maximized fully for his experiments and spikes.⁴¹

I’ve been using word count to try to explain why Ulysses ended up becoming as long as it did. To do so, I’ve been arguing that length—and its incremental changes over the course of two years—is intimately connected with the material constraints of serial production, the dilated space of the plot, and the unexpected exigencies of narrative experimentation. Now I’d like to entertain a related question: How long is an hour in Ulysses? Is a 5,432-word hour the same as one made up of 11,212 or 16,510 words? “The number of pages and lines in the published work,” Paul Ricoeur points out in Time and Narrative, are usually the measuring sticks for questions involving novel length and erzählte Zeit (chronological time): pages and lines, that is, not words.⁴² But with serial novels, where the length per installment was a very real consideration, one that often determined the design and tempo of a novel’s plot, word count was a primary, not a secondary, consideration. In Joyce’s case, where every episode was made to correspond with one hour of the day (or thereabouts), words on the page were like seconds on the clock, and still we know from the increase of words within and across these episodes that time was not always measured in the same way according to the same rubric.

Here, of course, we’re back to narrative, since this question of time and word count is bound up with Joyce’s experiments as he was writing. To get started, let’s make a baseline calculation using episodes 1 through 5, where, on average, one hour equals 5,233 words. In reading time that’s about 87 words per minute if it corresponds exactly with the time of the plot. Too slow. Average readers, capable of consuming between 250 and 300 words per minute, would be able to plow through an episode in about twenty minutes, though we would probably want to adjust for difficulty so that episode 3 and episode 2 would necessarily require different amounts of reading time. At this point, the calculation is pretty simple regardless of the ratio between interior monologue (interior/subjective) and third-person narration (exterior/objective). I say simple because the physical action in the external world of the novel and the thought in the interior worlds of Stephen and Bloom are equal in each episode and can be measured in the same way by the same clock. Individual thoughts pile up in the minds of Stephen and Bloom, and they mark the passing seconds as much as the commentary of the narrator does. “Stately, plump Buck Mulligan came from the stairhead”: that’s about five seconds, and we’re not even to the bowl of lather yet. More often than not thought and action are happening at the same time: Stephen walks down the beach as he’s thinking, he recites lines of William Blake in his head as the children are talking, Bloom oogles the woman in the butcher shop as he stands and waits in line, he reads the newspaper as he defecates, and he spreads the butter on his toast as he silently counts (“Another slice of bread and butter: three, four: right” [U, 45]).

But what happened to time in Ulysses when the word count per episode grew to more than 8,000; 11,000; and 19,000 words? Did the seconds slow down or speed up? Is an hour still an hour if there are more words on the page and in an episode? With the first spike from 5,000 or so words to 8,500, the answer is yes. In episodes 6, 7, and 8, more things happen within the space of an hour—more dialogue, more thoughts from Bloom, more social interaction among characters, more description of the environment—but time does not really change its pace, no slowing down or speeding up. A second, to put it another way, is still a second even if there are more words attached to the passing of each one. In the second spike from 9,000 to 11,500 that is not the case. Once these narrators get reintroduced with such strong personalities, they have a lot more to say. And not only that: it takes them longer to say it because they use more words. When Bloom and Stephen think (even at their wordiest), they can say less, a few words to make their point and on again (and for readers it becomes easier to recognize their thoughts because they revisit many of the same topics over and over again).43 But narrators need complete sentences that are considerably longer, and, as a result, they have a way of taking up more words while using fewer seconds so that the correlation between word and second is elongated.

More words on the page but fewer seconds passing in the plot: that is a discovery Joyce made while writing Ulysses. Time, expressed through words, is malleable, and the narrator is not bound to move it along with the description. That may help to explain why the length of the episodes needs to increase so dramatically during the middle stage when the narrator returns.44 Episode 9, after all, introduces the playful narrator who gets in the habit of not describing in any linear way: the sentences tend to slow down the action instead of moving it forward. And in episode 10, there is simultaneity, juxtaposition, and backtracking. The narrator in this instance describes different characters in different places, while the minutes of the clock are regularly pushed back to begin again. In addition, Joyce planted an interpolation within each of the nineteen sections marking events that are either happening, have happened, or will happen within that hour. Seen this way, episode 10 is a representation of the serial experience itself, a miniature of the entire novel as it appeared in installments, each one with its own pocket of time and space. Each section had to be consumed in a linear way, but the leaps back and forth across Dublin effectively shatter the assumption that time itself moves forward consistently just because the words are accumulating sentence by sentence.

And from episode 10 forward, time is never the same again. Episodes 12, 13, and 14 are the longest from Ulysses to appear in the Little Review, and all of them have dominant third-person narrators who describe only to delay the passing hour. There is a lot to say here about how each of them work individually, but for the sake of space (and time), I want to limit my point: Joyce is often credited, along with Virginia Woolf and Marcel Proust, with finding new ways to represent the subjective experience of time in the novel. But it was a discovery not limited to interior monologue alone. What Joyce discovered in the process of writing Ulysses was another way to make the time of third-person narration subjective, using the linearity of the sentence to slow down the clock. And none of it would have been possible if he hadn’t realized at some point in the composition process that the piling up of words could make time in the novel more elastic. Just think of that early sentence in episode 14 that kicks off the long parody of English prose styles that goes on for sixty-three paragraphs without any divisions between them. There are 118 words in that single sentence, and not one of them gives you the impression that the clock is ticking anywhere except in the room where you might be reading it. Sure, words may be lining up one after the other, but nothing seems to be happening at all, that is, nothing except the language, whose visual appearance on the page becomes the record of time itself. This process first began in episode 6, when Joyce started to move away from the assumption that time had to pass in thought or in action. By episode 14 thought becomes action but narration has become a mode of inaction, which may help to explain why these late episodes require “many more words,” as Groden points out, “to say much less.”⁴⁵

There’s another way to frame all of this: by learning to use more words, Joyce was able to break out from the realm of story into discourse. The plot is happening, events are unfolding over the course of the day, but the meaning of the novel is increasingly bound up with how they happen. Which is to say that once the technique begins to dominate, time becomes less rigid, and, more to the point, there is a way in which the sheer number of words in an episode could effectively make it seem that more time had passed than it had. Is it any surprise, then, to discover that Pound only began to complain about the length of Ulysses with the arrival of episode 11? In a letter dated June 10, 1919, Pound informed Joyce that episode 11 was “too long,” warning him that it might be a good idea to slow down with the delivery of the other episodes, presumably so they could be shortened.46 What’s so striking here is the fact that Pound put his finger on length in an effort to describe what was bothering him. But length wasn’t the problem. In fact, the previous two episodes were only slightly shorter. What Pound’s done, however, is elide length with experimentation. The experiment, as he sees it, goes on too long, and unlike the previous episodes, it has a way of making episode 11 seem like the wordiest hour to date. Pound’s advice, as we know, went entirely unheeded. Instead of slimming episodes down, Joyce continued to fatten them up so that by the time the typescript for episode 12 arrived at Pound’s door, it weighed in at a whopping 16,510 words, and was followed up four months later by a 14,231-word episode 13, with a 19,214-word episode 14 waiting just around the corner.⁴⁷

This frenzy for more words was justified by the formal experiments each episode required. The numbers could not have been anticipated from the start; they had to be discovered gradually, and with that middle stage especially, there was a way in which Joyce stopped caring about any limits, whether defined by words, lines, pages, or single/double/triple installments. And the point, finally, is that the Little Review always made the necessary accommodations so that Joyce never had to rein his novel in during the serialization process. What needs to be stressed above all is the idea that no matter how much Joyce used the exigencies of narrative technique as part of his own defense, none of it would have been possible if he had not made the discovery that more words were required. Without them episode 11 is less lyrical, episode 12 less windy, episode 13 less syrupy, and episode 14 less, well, embryonic. An hour may pass in every episode, but the experience of time is different in each, as they variously exercise the power to make the reader feel pulled along, drawn back, blown aside, rolled over, and—why not?—reborn.

Joyce’s lexicon was vast, but just how vast was it during the serialization process? More precisely, how were the different (appearing once in an episode) and unique (appearing only once in the entire work) words distributed over the course of two years, and what was the rate of expansion? The chart in figure 2.7 contains the number of different words per episode, showing that in the first five, the average starts somewhere around 1,750; rises to 2,500 in episodes 6 through 8; and jumps up to 3,400 in episodes 9 through 13. While the very idea of an increase in different words might not seem particularly compelling at first (the number of different words, after all, will increase with the number of total words), the percentages are. Word count, as one might expect, increases in these jumps between episodes 8 and 9, and episodes 11 and 12, but the percentage of different words to total words remains the same (29 to 30 percent; 26 percent and 26 percent). Episode 13 is on the lowest end at 22 percent, but the rest of them hover between 26 and 41 percent. As a point of comparison, consider the percentage of total to different words in Portrait of the Artist as a Young Man, where the word count for each of the five chapters varies in no particular order between 9,000 and 26,000. The percentage of different to total words in these chapters is, on average, 19 percent; that’s about 37 percent lower than Ulysses.

What’s happened? What does this percentage actually mean? Because there is no data set yet available that could let us compare different to total words on a large scale across thousands or tens of thousands of titles, it’s difficult to say with any certainty. However, if we compare Joyce to himself, a few provisional observations can be made. The most obvious is that he used more different words consistently per episode in Ulysses than he did per chapter in Portrait, where the percentages go from 13 to 22 to 19 to 25 to 18. So even as the number of total words increased, the percentage of different words remained roughly the same. This would mean that even if readers were encountering some of the same different words along the way, each episode would basically have a similar degree of word diversity. But what does that tell us about Ulysses? In a broader sense, it gives us a glimpse into the pacing process of the words so we can see what was happening to the language as episodes were coming out month by month. Joyce, of course, would not have been as conscious of word diversity as he might have been of word count, but that’s what makes these percentages all the more compelling: there’s a consistency over time, though, I admit, and it is one that still requires further analysis and broader contextualization if we want to understand how exceptional or ordinary it is in the universe of the novel.

The number of unique words within the different words is a better place to examine the lexical diversity of the serial Ulysses. It is already known that the number of unique words in the book is high (somewhere around 35,000), but what about the 9,580 unique words that appeared in the serial version? How were they distributed and where (figure 2.8)?⁴⁸ Episode 12 has the highest number of unique words (1,730) followed by episode 9 (1,282), episode 10 (941), episode 11 (936), episode 13 (870), episode 8 (720), and so on, all the way down to episode 5 (212). And it’s not just that there are more unique words in these middle-stage episodes (since, again, these middle episodes are longer, so an increase is to be expected): it’s that the percentage of unique words within the different words is actually higher (unlike the ratio of different words to total words, where the percentage stayed more or less the same): 41 percent for episode 12, 39 percent for episode 9, 31 percent for episode 11, 30 percent for episode 10, and 28 percent for episode 13. The middle-stage episodes average 34 percent, the early ones 24 percent. That difference suggests that the episodes in the middle stage weren’t just longer: they were getting filled up with a more expansive vocabulary.

Why? Let’s consider, for instance, the different percentages in the “Telemachia” of Stephen and the first three episodes of Bloom’s “Odyssey,” where the word count is between 16,760 and 17,849. In Stephen’s episodes, the average number of unique words to different words is 26 percent; in Bloom’s, the percentage drops down to 18 percent. The difference, I suspect, has a lot to do with who these characters are. Stephen’s braininess is actually reflected in the number of unique words he uses. And if you don’t believe me, take a look at the percentage in episode 9 where he delivers his lecture on Hamlet: at 39 percent it’s the second highest percentage in the entire list, right below episode 12, which is more than 5,000 words longer (with a unique word count that is higher by 500).

If Stephen uses more unique words than Bloom in the early episodes, it’s the narrator who trumps them both in the middle stage. I speculated earlier that the return of these third-person narrators was connected with the increase in word count, but in light of these percentages, I think it’s necessary to refine this statement a bit further: these narrators are not just using more words than Stephen or Leopold, they are using more unique words. Had the omniscient narrator never returned to the novel with such force, it is likely that the number of unique words would have stayed somewhere closer to the 22 percent range (as opposed to the 32.5 percent range, which was the average from episode 10 to episode 13). But that’s still higher than Portrait, which averages 18.4 percent of unique words for its five chapters. And who knows what else is out there waiting to be discovered by comparing Joyce with all of the great serial novelists of the nineteenth century. Ulysses is not a universe unto itself. Indeed, there are internal laws at work here, but understanding what they mean requires looking outward, situating this one “fat book” in a larger network that includes many, many more.

In the Poetics, Aristotle maintains that length is something that emerges organically from within the sequence of events of a particular play and should not be regulated by such things as the clepsydra, or “water-clock.”49 Ulysses proves to the contrary that the water clock—or in this case the serial installment—was indeed a powerful tool precisely because it set the pace for artistic production and influenced the scope and scale of the entire work of art. Once he heard the clock ticking, Joyce could no longer ignore the fact that the words were accumulating, time was passing, and the novel was growing. Along the way, these episodes were fast becoming part of a sequence, the length of each one determined in part by where it was placed. Indeed, if it was the space of adventure that first necessitated more words in episode 6, it was the potential for multiple installments that made it possible to experiment more fully with time and narration after episode 9. In the act of preparing each episode for the Little Review, Joyce found a way to transform story into discourse, making the hours seem longer when they were just getting wordier, and that, in the end, was one way for him to figure out what Ulysses could be.

Before closing, I want to make a few observations about the impact of serialization on the physical book published by Shakespeare & Company in February 1922. By the time the final 4,964-word installment appeared in the September–December 1920 issue of the Little Review, 121,864 words of Ulysses had been printed, and with the addition of four more episodes and a revision of the earlier ones in the next twenty-four months—where words were added and seldom subtracted—that total number more than doubled (figure 2.8).⁵⁰ Episode 15 is the most obvious indication of the impact that serialization had on the length of Ulysses: in the absence of deadlines and/or any material constraints, Joyce unleashed a 38,020-word monster. At this point in the process, he was writing for book publication, so length was no longer a consideration. But that point alone, one that Groden first made, should make apparent just how much the act of serialization on a monthly basis influenced the overall size of the book: it forced Joyce to figure out how many words per episode would be adequate for the story he wanted to tell and the techniques he eventually wanted to try out.⁵¹ And those word counts were not modified significantly for at least six of the Little Review episodes during the great revision of 1921, when Joyce was unifying the book by layering on all kinds of symbols and Homeric correspondences: 375 additional words for episode 1; 107 for episode 2; 261 for episode 3; 1,012 for episode 4; 871 for episode 9; 693 for episode 11; and 1,122 for episode 14.⁵² In their eagerness to downplay the relevance of the serial Ulysses on the basis of its textual mistakes, that’s not a point genetic critics want to make.⁵³ But it’s true. Even though Joyce went back to revise all of the Little Review episodes, six of them received only minor additions/substitutions averaging fewer than 1,000 words. And with the exception of episode 7, none of them received structural modifications: the order of the events was the same, the cast of characters all there (no one cut from the final program), the itineraries remained unmodified, and the hours of the day and all of the destinations stayed in place.⁵⁴

There’s one more thing to consider: even after serialization stopped, Joyce was still writing by numbers. In letter after letter, he bragged to friends about the massive size of episode 15.55 But even for him, 38,000 words was just too long for the three episodes he still had left to finish, and, by the time he came to put the finishing touches on episode 18, he was acutely aware of the fact that the first sentence was 2,500 words with seven “sesquipedalian” (literally “long-worded”) sentences remaining.⁵⁶ Should we really be surprised, then, to discover that episodes 16, 17, and 18 average 23,000 words and make up another sequence of three? Is 23,000 the ideal length for an episode without the burden of serial pressure? Maybe, but it could never have happened without that more modest beginning somewhere around 6,000, when the world of the novel was all before him and the rock of Ithaca far away on the horizon. But why 23,000? Well, that’s a riddle for someone else to solve, but it’s hard to believe that when getting these final episodes ready for the “book” book, which Pound would identify with “the size of four ordinary novels,” Joyce wasn’t thinking each of them would fit quite neatly in four installments.⁵⁷

Ulysses has been around long enough, but no one has yet tried to figure out something as simple as its word count, including the dimensions of the individual episodes as they were appearing serially. And this quantitative approach, as I have tried to demonstrate throughout this chapter, would have been meaningless had it not been complemented by the genetic backstory we now know so well. Groden’s work in the mid-1970s was groundbreaking, and it helped us see how a single novel could undergo so many significant structural changes over the course of a few years. In researching this topic, I was sometimes dazzled by how right Groden turned out to be, reaching his conclusions using so many different notes, notebooks, manuscripts, typescripts, and proofs. But there are limits here as well, and sifting through the genetic evidence can only tell us so much. In order to try to explain why the episodes got as long as they did, it was necessary to read Ulysses by numbers. By doing so, it became possible to generate a data set that revealed variations, or spikes, thereby requiring an explanation that could, whenever possible, try to account for them.