Ulysses by Numbers

Who wrote Ulysses? Probably not the question you were expecting at this point, but before responding (or dismissing it entirely), consider what evidence you’ll draw on. All of it will be based on the firsthand or secondhand testimony from friends, colleagues, and acquaintances who knew the person responsible for creating it; then there will be the estimation of critics, some of who have worked closely with the relevant authenticated documents. This kind of authorial proof was not available to Homer, of course, and very little of it was around for Shakespeare, a writer still hounded by suspicions of authenticity.1 Still, we know enough about the worlds in which they lived that we feel comfortable identifying the works they composed with distinct styles and personalities (even if for Homer it is a collective one). Questions about authorship are significantly less common for modern authors—in part, because they tend to leave so much material behind, and eye-witness testimony abounds. Joyce, the one born on February 2, 1882 in Dublin, is among them: his major works at the center of a vast constellation of secondary sources. But before getting sidetracked by how difficult it may be not to know an author nowadays, let’s return to that original question: I’m not asking who is James Joyce or even what is Ulysses. Instead, I want to know who wrote Ulysses.

Style is the category that gets used to try and answer such questions, though the approaches vary widely. There is no single way to prove how style is the man or woman, and this is especially true when we’re talking about literary style. Joyce can sound like Joyce, yet it’s not always clear what the source of the singularity may be: word choice, sentence structure, characterization, modes of narration, or even subject matter. If following the experiments of Harry Mathews, you were to take the structure of Ulysses and insert a different plot with different words, would it still sound like Joyce?² Or what if you kept the same words and plot but then added a different structure? When it comes to literary style, it may not be possible to boil it all down to a single element, yet, what is true for this novel is true for so many of the others we read. Something about Ulysses sounds like Joyce and vice versa. Part of his style is unconscious, shaped by the period in which he was writing, other parts intentional, emerging out of a deliberate interest to use language and syntax differently. In any case, style is not biographical, by which I mean we cannot pin it down to an event in Joyce’s life or to the circumstances of his upbringing. Style is something else. It is infused into the writing, making it possible to sound authentic even when all the words are borrowed.

Conspiracy theories about the authorship of Ulysses do exist, including one claim that Ettore Schmitz (a.k.a. Italo Svevo) is behind it all. But in the 1960s, the study of literary style was fast becoming a subject of immense interest for statisticians thanks to the arrival of computers that could analyze lexical diversity on a large corpus of texts and across a wide range of genres and authors.3 One of the more successful early examples involved the Federalist Papers written by Alexander Hamilton, James Madison, and John Jay between October 1787 and May 1788. A majority of these pamphlets were published anonymously under the name “Publius,” inspiring two statisticians (Frederick Mosteller and David Wallace) to examine the frequency of noncontextual words (conjunctions, prepositions, articles such as upon, enough) and nonfunctional words (by, of, to). Once they determined the thirty words that most clearly distinguished Madison and Hamilton from one another, they built a statistical model and plugged the code into an IBM 7090 computer three thousand words at a time. Relying on Bayes’ theorem, which calculates probability based on previous events, they were able to solve this stylistic whodunit: Madison, they determined, is the principal author.⁴ The bigger takeaway, however, was the proof that the computer-assisted statistical analysis of lexical diversity could be used effectively to authenticate literary identity.

Not all of the computer-assisted stylistic experiments in the early 1960s were as successful. A. Q. Morton, a Scottish minister, used computer analysis to decode the style of St. Paul, leading him to the conclusion that only four of the fourteen epistles can be attributed to him with any certainty (the others belong to five other hands).⁵ The problem wasn’t the computer: it was Morton’s method. By focusing on such things as first and last words in the sentences (a technique he called “positional stylometry”), he didn’t account for the fact that Greek manuscripts in prose tended not to have marks of punctuation, so the whole idea of where a sentence begins and ends can be difficult to establish. One of the most scathing responses came from John W. Ellison, already established for his work producing a concordance to the Revised Standard version of the Bible with a RAND Univac I computer, who threatened to use Morton’s own method against him.⁶ But in order to show just how misguided Morton was about St. Paul (even claiming it an abuse of scholarship and computers), Ellison brought Ulysses into the argument. A New York Times columnist, who was among the room full of conference goers at Yale University, reports: “The rector said that by using the method, he thought he could ‘prove that five authors wrote James Joyce’s Ulysses,’ and that none of them wrote Portrait of the Artist as a Young Man. At least six authors could thereby be assigned to the two novels.”⁷

Five authors, one Ulysses, and no Portrait: if it were possible to transport Ulysses to a world where Joyce was entirely unknown, that conclusion might not seem so outlandish. Still, there are a few other points worth making. For one, Ellison never actually performed any stylistic analysis on Ulysses with a computer (probably because of the amount of labor involved), so his comment is entirely hypothetical, simply meant to foreground the absurdity of ever thinking that literary style could be measured by sentence length and positional stylometry. The second, however, is that he uses this particular novel to make the point. Ellison’s joke is funny because it assumes that Ulysses could only have been written by one person, and the mere thought of suggesting otherwise would be an insult not just to Joyce and his readers but also to stylometry.

Ellison may have dismissed sentence length as a reliable indicator of style, even if it was a hot topic for literary-minded statisticians in the precomputer days of the 1940s, but more recently a group of nuclear physicists used it as the point of analysis for a corpus of 113 literary works in English, French, German, Italian, Polish, Russian, and Spanish.8 Their conclusion: Finnegans Wake is a multifractal (fractal of a fractal), which means the word counts within sixteen thousand sentences form a consistent pattern across the work as a whole (figure 1.1).⁹ More surprising, out of this sample Finnegans Wake was the closest to a mathematically pure fractal, more so than Woolf’s The Waves, John Dos Passos’s U.S.A. Trilogy, and Roberto Bolaño’s 2666. Why a group of nuclear physicists would be working on this particular problem is unclear, but their conclusion reveals that multifractal structures can, indeed, be a useful indicator of style. It’s not the content or diversity of the words that tip them off: it’s the variation in sentence length (more than sixteen thousand) across an entire work.¹⁰

Strange to think that in all this time, and with so many different studies of styles and languages, no one has ever tried to prove by computer who wrote Ulysses.¹¹ The most ambitious computer-assisted experiment so far involved Gabler and his specially designed TUSTEP program, but that was for the production of an edition, and the handlist Steppe and Gabler generated was more of an alphabetized afterthought intended for word searches. There’s still time, but I make this observation if only to highlight that the style of Ulysses seems so self-evident, so obvious, so basic, so singular that no machine and no statistical or computational model could ever teach us things we don’t know already.

But for the sake of Ulysses, and the potential value of literary computational methods more generally, what would happen if we were to take this question seriously and approach it as a novel without an author? What would we choose to focus on if we wanted to examine style quantitatively? In the century or so that stylometry has been around, units for analysis have included the length and frequency of words and sentences, and they work rather well with models involving statistical probability. Authors become identifiable not only by word choice but also by word adjacency and lexical diversity.

Another option, one foundational to the organization any literary structure in prose, will involve the one element that few, if any of us, take much notice of: the paragraph.12 As unmemorable as the paragraph may seem, it works behind the scenes to keep all of the words and sentences in order. By using the term order, I mean to suggest that the paragraph has a design, it does not just appear because of a predetermined decision made by a machine. Instead, it is chosen by the author, left there to perform a great deal of semantic work, bringing ideas together and also breaking them apart. In addition, the paragraph is there to be counted, and anyone who pays attention begins to realize just how formative the paragraph is not only to the pacing of an argument or plot but also to the arrangement on the page. Having too few or too many paragraphs is something readers will notice, but it’s that in-betweenness, often a combination of short and long, that can slip by without much fanfare. Still, paragraphs, though extraordinarily common, actually do something, focusing attention with an enforced pause or straining attention by withholding one, thereby forcing readers to hang on as the sentences pile up.

If the paragraph can have a style, it is also in the quantity, one that gets defined by the number of starts and stops that an author decides upon (figure 1.2).¹³ But if everyone agrees that the paragraph can be defined by what it does, there is no consensus on how often this process should occur or whether the variation in size or sum even matters. Gertrude Stein seemed to think so. Even more than the sentence, paragraphs, she argues, have an “emotional” content not because of what they say, but rather because of how “they register or limit an emotion.”¹⁴ That limit is precisely where they derive their power: the paragraph, in other words, has to begin and end (usually in a sequence), but no matter how long it may wander, this unit is finite. It’s the finitude, in fact, that makes the paragraph, that “understudied scale of prose writing,” measurable from a wide variety of angles, some of them I’ll address in this chapter.¹⁵

But if paragraphs can have a thematic or material limit, they also have the power to ground narrative experiments so that they don’t short-circuit by running on for too long. So no matter how surprising the words, sentences, or episode structures become, the paragraph remains. Not even Molly Bloom could kill it off with her extended monologue or the headline-obsessed narrator of episode 7 or the cheeky narrator from episode 10 with its sudden temporal and spatial displacements separated by asterisks. And if Joyce found himself at the end of the English language with Ulysses, he held on to that paragraph for dear life and never let go even after surrendering to the dreamscape of Finnegans Wake, which begins in the middle of a sentence that culminates in—what else?—a final incomplete paragraph.

Powerful or not, the paragraph is still one of the more overlooked organizational units in Ulysses. This has not been the case for Marcel Proust. Fredric Jameson, for example, has called the paragraphs from À la recherche Proust’s “most striking innovation,” and that’s for a writer who brought the madeleine, and by extension the flood of memory, to the modern novel. Part of this innovation comes from the fact that Proust was using paragraphs to “organize human experience,” inserting so much material in them during the revision process that Jameson wondered “how, under such conditions, any sense of completeness or totality is possible.”¹⁶ The answer, as Jameson knows, is irrelevant. There is no totality in the Proustian paragraph because the content of the memory as documented on the page always exceeds the act of remembering. The paragraph, then, is fundamental to the construction of a memory-theater put there to demonstrate the impossibility of ever fully capturing the human experience with an ever-receding time.

Joyce’s paragraphs may not be as monumental an innovation, but they were certainly an innovation. When consulting the piles of manuscripts and drafts, Jean-Michel Rabaté and Daniel Ferrer noticed that Joyce, like Proust, had a tendency to stuff his paragraphs with words, sentences, and phrases during the revision process, but it was seldom the case that he added new ones.17 These so-called paragraphs in expansion not only meant the size of the novel would increase, but also there was the implication that the process could continue so long as there was time left, the finishedness of the paragraph only imposed by the necessity of publishing deadlines and death. The paragraph remains the perfect unit for anyone eager to take a close look at Joyce’s creative process, and depending on which example you choose (and the number of revisions that followed), the results can be pretty revealing.

The other side of the genetic equation, however, deserves our attention as well. What do we make of the fact that paragraphs expand in size during the revision process but rarely in quantity? For some episodes, the answer to this question will depend on where the episode falls on the genetic timeline. If the earlier episodes, once proofed, were more or less off limits to further revision, the later ones, still in process, were an opportunity to weave in recurring words, themes, and motifs. That, Kenner suspects, is one reason why the novel gets more complicated as it goes: later episodes were more available to tinkering, the kind that tends to happen inside the paragraph.¹⁸ As a general rule, the quantity in the episodes remained more or less stable.

Such was the case, for instance, with episode 12, which went through nine different proof stages. Michael Groden noticed that the encyclopedism of this episode was the result of late-stage additions, but with only “two entirely new paragraphs added.”¹⁹ The effect is significant. “The additions,” Groden argues, “change the episode’s pace, as they give the passages of gigantism an increased life of their own.”²⁰ And then there’s episode 14. This is the one with sixty-three paragraphs that are meant to correspond with the arrival of the sperm to the egg, the weeks of gestation, and the exit of the baby from the womb. Given such a rigidly defined structure, there was really no room for new paragraphs, but the number, for those who counted, was evidence that the quantity was integral to the overall effect and meaning of the performance. Add one or two more paragraphs to episode 14, and the correspondence between language, birth, history, time, and creation collapses—even for the readers who never kept track.

And here we are, already sixteen paragraphs into this chapter, and I still haven’t even defined the term. In the Oxford English Dictionary, a paragraph is defined as “a distinct passage or section of a text, usually composed of several sentences, dealing with a particular point, a short episode in a narrative, a single piece of direct speech, etc.”21 No surprises here, but consider all of the quantifying that the paragraph brings with it: several, particular, short, single. In order to define the paragraph, it becomes necessary to imagine how it can be distinct, something that can be extracted from a larger body of text, and still connected, something comprised of sentences and parts of speech. The point is that a paragraph can have its own identity, but that doesn’t mean it can, or even should, exist on its own. So while it might be possible to condense the stories “worth writing” into a paragraph, William Faulkner, at least, was someone unwilling to recommend one-paragraph stories.²²

Before deciding whether or not paragraphs can even have style, we should be clear about what we mean by paragraph.²³ The Oxford English Dictionary definition fails to indicate that the concept of the paragraph was around long before moveable type. In ancient Greek manuscripts, paragraphos identified interlinear markings made in the left margin to signal a change in speaker.²⁴ As this typographical convention circulated over the centuries, scribes developed other symbols that could be used to mark a new speaker or topic (K for kaput, or head), and it came to be called a pilcrow. Eventually, scribes would keep the space of pilcrow blank so that rubricators, the ones in charge of inserting the first letters, could provide elaborate designs by hand. With the arrival of the printing press, scribes and rubricators alike were left with less time to complete their tasks, and before long the paragraph was transformed from a stylized symbol in the margins into a blank space. By the nineteenth century, the distance of that blank space was standardized, and it would measure one em dash.²⁵ And so it is that the paragraph defined as a line break followed by an indent with a capitalized first sentence is still with us today both in print and on screen.

If only things were that simple. “The internal length and organization vary infinitely” [“Sa longueur et son organization internes variant à l’infini”], one theorist of the paragraph writes.26 The paragraph may be a printer’s convention, but it is the writers who get to decide not only on length and internal organization but also frequency. Joyce was among them. Ulysses, then, begins like any other novel with an indented first line, and if you follow the sequence of the next few sentences, it ends with a colon, followed by an indented em dash preceding a Latin phrase, and then another indented line, another colon and another em dash accompanied by a line of dialogue. On the page, it looks like this:

Seven lines of Ulysses with four indented sentences. Nothing particularly unique here, but this example raises an important question, and it is one that will affect the analysis that follows: Does an indent always signify a paragraph break? You can see how it makes sense for the first line, but why is the dialogue indented as well and then followed by an indented single line in third person? Can a line of dialogue be a paragraph? And what about a single line of third-person narration? Is that a paragraph as well?

In other words, what counts as a paragraph? If it’s the indent, then any line of dialogue can fit the bill, but it’s even more complicated if you consider the way Joyce formatted dialogue in his manuscript drafts, putting each dialogue dash flush left so that it was in line with the unindented part of the preceding paragraph. In order to format the computerized version of Ulysses in the 1980s, Gabler needed those line breaks for dialogue, using the same command to create separate narrative paragraphs. In preparing the extensible markup language (XML) version though, he decided to deparagraph. Here is what you find (hint: <p>= paragraph):

What was once four paragraphs is now two. The <p>, also known as the paragraph element in HTML, is exactly where it should be in the first line—that is, before the indented narrative paragraph. You’ll notice, however, that the lines of dialogue are also included. The deparagraphing has combined the dialogue with the paragraph preceding it.27 That is not how it appeared in the 1922 version or so many of the editions that followed (Gabler excepted).

To paragraph or not to paragraph may not seem particularly urgent, especially for those readers who prefer to imagine this unit as a convenient prose container. But it’s definitely an issue for anyone concerned with the relationship between narrative frames, which includes those blocks of text organized by paragraphs and the extranarrative elements (including dialogue but also snippets of song and other bits of text) that often follow in between. Once you begin paying attention both to the structure and quantity of paragraphs, it makes perfect sense. The line of dialogue is an extension of the narrative preceding it but not automatically the one that follows. Take, for example, the moment when Hynes walks up to Bloom at Dignam’s funeral. “No: coming to me,” Bloom thinks before a line break followed by a few lines of dialogue: “I am just taking the names, Hynes said below his breath.” The conversation continues, but breaks again into another paragraph, this time about M’Coy (U, 92). We may clearly see the blocks of dialogue on the page but, they are not a self-contained paragraph. With few exceptions, they follow from the paragraph that precedes it.

With interior monologue, things will get a lot more complicated, but for now, I want to keep the focus on that leap between line break and indent, one we barely register as we make our way from the top of the page to the bottom. Buck Mulligan holding up his bowl and intoning Introibo ad altare Dei is part of a sequence: first the bowl, then the words. And this formula gets repeated more times than anyone would probably want to count. But then consider what this tells us about the paragraph. For one thing, it is not defined by the indent alone: it contains spoken words that are still technically placed outside of it. To define the paragraph in this way, then, requires acknowledging that it is an elastic unit, one with the capacity to incorporate words that are, strictly speaking, external to it.

But if adding dialogue to the preceding narrative can help keep paragraphs organized by topic, there’s one problem: it’s hard to know exactly how to count. For Kidd, every line of dialogue represented a paragraph, and using this definition, he counted one hundred of them in episode 3, which, he believed, corresponded with the one hundred cantos in the three canticles of Dante’s Commedia. What’s more, he singled out paragraph 33 for containing the phrase “The grandest number…in the whole opera” and paragraph 100 for its references to a “threemaster ship” and “crosstrees,” both of them, he argued, allusions to Christian symbols. In Kidd’s reading, the paragraph can be symbolic in its quantity.28 So as much as we know about numerology’s deep roots in words, sentences, sections, lines, chapters, and pages, never before has the paragraph been among them. But here it is: a sum of paragraphs with epic ambitions and mystical intentions.

And maybe this particular example shouldn’t be so surprising given that Joyce got into the game himself with episode 14. As Kenner explains, this episode opens “with eleven paragraphs that get Bloom into the Mater. It closes with eleven paragraphs outside again. Eleven is Joyce’s inception-number, re-creation number. And between this pair of elevens, inside the Mater, we may count exactly forty paragraphs: forty for the forty weeks of gestation.”29 And then there’s another example in episode 18. Though critics tend to refer to the sentences of Molly’s soliloquy, they are ignoring the indents. Episode 18 is actually comprised of eight paragraphs made up of two sentences (marked by two periods), and if we’re willing to factor in a reference to this episode from Finnegans Wake, that’s how Joyce came to think of it too: “thank Maurice [Darantière], lastly when all is zed and done, the penelopean patience of its last paraphe.”³⁰ Paraphe is the French word for “initials” or “signature,” but it’s also a reference to the very unit upon which the episode was built. And if the number 8 is a heavily charged symbol, invoking the day of her birth and the infinity sign, it’s the paragraph that set the limit, giving episode 18 a structure that changes topic and pace even if it’s only the readers in Molly’s head who would notice.

“Paragraphs contained style but they did not shape it.” That’s what the Stanford Literary Lab, following the paragraph-oriented criticism of Erich Auerbach and Ian Watt, concluded in their own work on the subject.³¹ But if we take into account these three examples, isn’t Ulysses an exception? Aren’t these paragraphs doing more than containing style? Though the dimensions of the individual paragraph may change, it’s the quantity that defines the content over the course of an entire episode. The significance of what’s happening inside the paragraph, then, is dependent on the sum (100, 63, 8), and readers who do not know the number miss the symbolism that holds it all together. In saying that, of course, I realize that each of these symbolic correspondences can be stretched to the realm of absurdity—yet, they are still foundational to the overall structure as it exists in this novel. What’s more, they provide different answers to the question: How many paragraphs are enough for an episode? It depends on what the content demands. And in this eccentric system of signification, sixty-three paragraphs can mirror embryological development, eight paragraphs can signal eternity, and one hundred can contain an entire cosmos.

Consider, though, how these examples would also complicate any computational project that uses big data sets to arrive at broader conclusions. These counts might tell you how the paragraph works in Ulysses, but they would not be particularly useful for anyone working across a much larger sample with hundreds or thousands of novels. I bring this up now to emphasize the specificity of the unit itself, especially given that it is less than likely to attract much attention. And if using the paragraph to isolate a few lines of prose or break up a thought is one thing, assembling an entire collection of them in order to structure an episode is quite another. Paragraphs do have style once you figure out how to count them.

Counting paragraphs brings its own set of complications. That has a lot to do with Joyce’s experiments with page layout, punctuation, em dash dialogue, and the inclusion of extraliterary material (including the printed matter in newspapers, songs, advertisements, etc.). In the process of trying to keep these numbers as straight as possible, I have established a few rules (with exceptions): the paragraph must have an indent (but not all lines with indents are necessarily paragraphs); dialogue is counted along with the body of the main paragraph preceding it (as was laid out in the original 1922 edition); bits of songs or remembered quotations are generally not included as separate paragraphs. A handful of episodes still required more inventive solutions: I count the 59 lines of the overture in episode 11, for instance, as one paragraph; in episode 15, the paragraph unit includes speaker plus stage directions and dialogue; in episode 17, each “paragraph” is a combination of question and answer. Some may not be entirely satisfied with these rules (and their occasional exceptions), but the main point of the chapter still stands: the standard definition of a paragraph as a unit marked off by an indent doesn’t work if you are interested in exploring larger questions involving the overall proportion and design. The quantities and dimensions of the paragraphs were made to accommodate a whole range of experiments with time, space, narrative, character, and, in general terms, world-building.

To try and answer the question, we need to expand the category to accommodate different ways of measuring paragraph size and structure. In the most comprehensive book on the English prose paragraph dating back to 1894, E. H. Lewis uses quantitative measurements to explain how the paragraph has evolved over the course of three centuries, focusing on such things as average sentence length and the average number of words per paragraph and per sentence. In the process, he discovers that those long, single-sentence paragraphs so beloved by Bunyan, Sterne, and Defoe fall out of fashion (he suspects because sentences get shorter), but the turn toward shorter sentences, which coincides with a decrease in conjunctions, also ends up reducing the average length of paragraphs. It is in the relative differences of paragraph size (including sentence length), however, that Lewis identifies a style that is understood collectively.32 An author, in other words, can be singular, but he or she is using paragraphs in a way that he identifies with a trend at a specific moment in time and place.

Lewis draws on a wide variety of prosaists for examples but also includes one novel by Charles Dickens and another by George Eliot for comparison. In a selection of 300 paragraphs (comprising 15,202 words) taken from Old Curiosity Shop, Lewis discovers the “coherence” of an “oral style,” and if, at times, the sentences can be “diffuse,” they are never “obscure.” But if Dickens has an ear for the sound of words and a tendency to let his sentences ramble for dramatic effect, Lewis concludes he “has no right feeling for the paragraph as a rhythmic whole.”³³ Eliot, on the contrary, does. The 212 paragraphs (16,233 words) sampled from Daniel Deronda are similar in size to Old Curiosity Shop, but they have an overall “unity” and “logical coherence” that Dickens lacks.³⁴ And while these stylistic differences may be noticeable for anyone who actually reads what’s inside the paragraphs, Lewis attributes them to the way paragraphs are organized, which involves, among other things, the timing and placing of short sentences for rhetorical effect.

There are a few important takeaways from Lewis’s book: first, the paragraph as a unit is measurable; second, measuring paragraphs, both individually and in aggregate, tells us about how a writer chooses to organize the narrative; third, and most importantly, the English paragraph itself has a history that involves changes not just to size but to the arrangement of the contents. In their analysis of the paragraph a century later, the Stanford Literary Lab gave Lewis’s quantitative methodology an update. Instead of counting paragraphs and sentences manually, they rely on topic modeling, using samples of narrative paragraphs, not including dialogue, for their experiments. Their discovery: the paragraph of the nineteenth-century novel prefers polytopicality. Instead of restricting themes to a quantity of one per paragraph, it was more common to find two, three, or four—but what this means, in the bigger picture, is that the paragraph provides access to the novel at a different scale. It is at the scale of the paragraph, they conclude, that we see how themes in the novel fully exist.35

Ulysses wasn’t included in either dataset. Lewis was looking for broad trends across the centuries before Ulysses was even written, while the lab focused its energies on the internal structure in the nineteenth-century novel. That is not the goal in what follows. Instead of looking to draw broad conclusions about the history of the paragraph, I use the numbers to explore the proportion and design in this novel. In particular, I am interested not only in how the paragraph develops across episodes but also in relation to one another as the size of the novel increases. As was true with words and characters, the quantity of paragraphs per episode was never a given. It would change as the writing of Ulysses progressed, responding to a whole range of symbolic filters and also experiments with language and narrative that were shaping how the plot was organized.

Joyce knew how to use paragraphs long before Ulysses, but one thing, in particular, changed it all: interior monologue. More than a technique for capturing consciousness, interior monologue was also a style of writing defined by distinct limits at the scale of the word, sentence, paragraph, and page. Adrienne Monnier was quick to notice the “short, cut-up, exaggeratedly allusive little sentences” of interior monologue but failed to mention that they were collected in paragraphs, some providing smooth transitions, others more abrupt nonsequiturs.³⁶ Inside and outside, interior monologue and third person, thought and action: that’s the shift that happens between line break and indent. The paragraph is put there to end and begin over and over again, not only interrupting thoughts but also moving the plot forward but not always with a clear connection.

But more than performing a convenient narrative function, the paragraph is, to borrow S. L. Goldberg’s phrase, a “dramatic unit of consciousness,” a material reminder that no matter how much Stephen or Bloom have their heads in the clouds, there is still a page in the outside world that needs navigating.37 But what happens when the paragraph becomes a “dramatic unit of consciousness” and is not, strictly speaking, a typographical convention for stacking sentences? What, then, is the paragraph counting, and how might we measure it for ourselves? Here, I want to identify the criticism leveled against Ulysses by Wyndham Lewis regarding this new narrative technique: “He had to pretend that we were really surprising the private thought of a real and average human creature, Mr. Bloom. But the fact is that Mr. Bloom was abnormally wordy. He thought in words, not images, for our benefit, in a fashion as unreal from the point of view of the strictest naturalist dogma as a Hamlet soliloquy.”³⁸ It’s the wordiness of interior monologue that irritates Lewis, but the emphasis could just as easily be placed elsewhere: Bloom is abnormally paragraphy. If the punctuated words strain the limits of disbelief required for fiction, then so too does the paragraph, which all too often takes more time to read than the experience of the thought itself. And, really, who thinks in paragraphs?

Marcel Proust let his narrator’s mind wander extensively, delaying the line break in order to simulate the whimsy of memory movement, but that’s not how things work in Ulysses. In fact, you could say that Joyce has a tendency to overregulate the paragraph to keep the minds of Bloom and Stephen in check, making them function in the novel much like a word bubble might in the world of comics. The text of his character’s thoughts is contained within a structure that has definite limits on the page. So as much as paragraphs organize, arrange, and pace consciousness, they also visualize it. Theodor Adorno identified something similar happening with punctuation marks, but for him it was a process dependent upon resemblance. “An exclamation point,” he writes, “looks like an index finger raised in warning; a question mark looks like a flashgun light or the blink of an eye.”³⁹ In Ulysses, the paragraphs don’t look like anything. They are blocks of words in various sizes with line breaks and indents. Once you try to imagine this novel without them, then you also begin to realize just how much they do to begin and end thoughts and provide necessary interruptions to bring in the sounds, smells, and sights of the outside world.

One of the earliest examples of paragraphed interior monologue is found a few pages into the first episode. As Buck Mulligan walks down the stairs into the Martello Tower after shaving, Stephen is left to think in peace:

Pretty standard stuff when it comes to interior monologue, but consider the role that the paragraph plays in all of this. Here, the indent is a signal, a term Wolfgang Iser uses to identify how modernist writers, in particular, can lead readers through “different depths of consciousness” without needing to rely on “extraneous codes.”40 In the example above, the indent is the signal identifying that a switch is being made, which in this instance involves the move from third to first person without the necessity of adding a “he thought,” or, more appropriate to Stephen, a “he brooded.”

And this move from one paragraph to another, from one mode of narration to another, is more than a signal, it’s an image. One thing that happens between Portrait and Ulysses is the discovery of what prosy thought can look like—not just as an assemblage of free-floating words of different sizes at different angles (what the Italian futurists called parole in libertà) but as blocks. Readers, then, don’t simply wade through the short, choppy sentences of interior monologue in an unmediated fashion: they see them.41 The line break, followed by the indent, is there to let everyone know that a change is about to take place. It’s a common syntactical feature even in the episodes where there is a lot of shifting back and forth between the narrative registers. Interior monologue had to assume a form on the page, and it is one that will necessarily change the more that Joyce realizes that paragraphs are not just collections of sentences: they have the power to make thought visible.

There are limits, of course, to the shape that these images can take. They are still comprised of words; but just glancing at a few pages, you can see the variations generated by the different positions of line breaks and the quantity of sentences. Monnier, then, may have noticed the choppiness of interior monologue at the level of the word and line, but she failed to mention that they were accompanied from the early stage by the paragraph. Here’s a rapid sequence of three, appearing after the pair of long paragraphs with which episode 3 opens:

Rhythm begins, you see. I hear. A catalectic tetrameter of iambs marching. No, agallop: deline the mare.

Open your eyes now. I will. One moment. Has all vanished since? If I open and am for ever in the black adiaphane. Basta! I will see if I can see.

See now. There all the time without you: and ever shall be, world without end (U, 31).

Difficult to pin down what is happening. If we are supposed to be inside Stephen’s head, what are the paragraphs doing here? Who is breaking these lines? And why are they so short? If the paragraph is, indeed, an image, it is not mimetic. Thought doesn’t really look like that in anyone’s head, but it can on the page of a novel. Joyce is clearly experimenting with the rhetorical possibilities of the paragraph as a unit. It does not have to follow the same logic that guides third-person narrators and can serve a semantic function, one that informs how readers process the words. Stephen may think in prose, but his thoughts are clearly guided by the sound and rhythm of poetry.

Even this one early example demonstrates that Joyce is adapting a technique developed by Stéphane Mallarmé, the symbolist who used short paragraphs surrounded by blank space as a display case on the page for his poetic prose. His justification for this practice could just as easily be a description for the three short paragraphs above:

Thought in the form of prosody contained in a paragraph: there’s no better way to describe what’s happening here. But more than marking thought, these paragraphs are also up to something else: they measure time.⁴³ Each episode lasts an hour, which would mean that the paragraph, like the word, must correspond with the passing seconds and minutes. And if the narration has a clock ticking, with different paragraph lengths, then it would be worth thinking about whether or not quantities of paragraphs conceal quantities of time.

If we restrict our analysis for now to episode 3, we are dealing with seventy-five total paragraphs.44 These seventy-five paragraphs contain all of Stephen’s thoughts and are largely organized around three different topics. These include an imaginary visit to his Uncle Richie Goulding; a real, albeit remembered, visit to Kevin Egan in Paris; and a long-distance encounter with the cocklepickers on the beach. The paragraph, though, doesn’t just provide the foundation for memory-building: it is part of the visual expression of the performance, with the bigger units, sixty, conveying the bulk of the memories and punctuated now and again by fifteen others made up of two lines or less.

If we want to understand how time is measured, the short paragraphs are a good place to start. While the bigger paragraphs tend to get organized by topic, concept, or event (as happening in present or as remembered in the past), the shorter ones either mark a quick change in topic/concept/idea/event (“What about what? What else were they invented for?”), or they act as the dramatic culmination of a thought or memory (“Weak wasting hand on mine. They have forgotten Kevin Egan, not he them. Remembering thee, O Sion”), or they provide an interruption from the external world (“Passing Now”) (U, 34, 37, 39). But there’s something else going on as well: the words contained in these shorter paragraphs are in step with Stephen. As he walks and thinks, the seconds tick by, and the short paragraphs, distinct on the page, reveal a self-consciousness about this connection between his body and mind in space and time.

Consider, for example, this line: “The Ship, half twelve. By the way go easy with that money like a good young imbecile. Yes I must” (U, 32). Here we have a thought about Mulligan, followed immediately in the next paragraph with the information that “His pace slackened.” Another longer paragraph culminates with “Ringsend: wigwams of brown steersmen and master mariners. Human shells” with a line break, an indent, and a “He halted.” Yet another involving memories of Paris (“You seem to have enjoyed yourself”) begin in the next paragraph with “Proudly walking.” These examples indicate that the narration, no matter how windy, is keeping track of Stephen. In that way, he is not unlike Dante, whose passi, or steps, in the Inferno, are both literal (Dante the pilgrim is walking) and figurative (Dante the poet is writing the lines we read). He walks, circle by circle, but no matter how deep he goes, Dante, like Joyce, never forgets the time or the pace.

Long paragraphs, which move away from the external world and physical action, tend to work differently. The hour is still passing in the episode but at a different rate. Instead of real time measured rhythmically and by an awareness of the body in space, the longer paragraph functions more like a pocket of time, a place in which the seconds and minutes can expand or slow down. And that’s largely because thought is not action, so measuring the time of thought, especially realized as words or in the form of sentences, would end up contradicting the process. Stephen, then, is free to dwell in the realm of abstractions in these long paragraphs so even when remembered or imagined actions enter his mind (talking to his uncle or Kevin Egan, for example), they become part of his network of ideas but do not have to be anchored in time so quickly.

Sixty paragraphs and sixty minutes is not so surprising. Out of all the possible things to worry about in episode 3, the shape of time, of course, is paramount. An innovation like interior monologue still needs a stopwatch, since the very realism of the novel demanded it. Stephen, after all, is in a real place, with a real body, moving through a fixed distance. So regardless of how oblivious he may be about the passing minutes, the structure itself, the one holding the sentences all together, has a number whose meaning is temporally significant. To measure time by sixty, in this episode and anywhere, is to adopt the sexagesimal numeral system invented by the Sumerians, and it is suspected that this number was selected, in part, because it is the smallest number divisible by the first six counting numbers.45

Stephen wouldn’t know any of this, of course, but the paragraph, which contains his words and structures his thoughts, has a temporal measurement. In the last section of this chapter, I’ll return to this question about numbers, paragraphs, and time; but for now, I want to make a simple point: we don’t need intention to make sixty paragraphs legitimate. Joyce may or may not have known that they could be added up in this manner. Regardless, he is still managing proportion, and this episode presents a unique challenge because it was the first but not the last episode that showcased interior monologue exclusively. But if we also bear in mind that this only happens again with Molly’s eight paragraphs, which begin in bed and end in eternity, the sixty could represent a different example of mind-time measured in minutes. The minutes of episode 3 may well have been a dress rehearsal for the eternity of episode 18, one that led to the realization not only that a quantity of paragraphs could be symbolic, but they could also be adaptable for experiments with form and temporality.

This is the paragraph measurement for a single episode, but even if Joyce measures all the other ones by the hour, the quantities will vary. Take, for example, the next set of three, which accompany Bloom’s arrival in the plot: 107 (episode 4) to 87 (episode 5) to 197 (episode 6). Episode 6 is on the high end, but that makes sense given the increase in the word count (8,388). And for many of the episodes that follow, a similar process unfolds: more words=more paragraphs.46 That in itself is not very surprising—especially when you acknowledge the return of third-person narrators who not only use longer sentences but also require more paragraphs to fit them in. It’s the reverse trend, however, that requires more scrutiny. With its twenty thousand words and sixty-three paragraphs, episode 14 is an obvious exception to the more words/more paragraphs rule. Next to an eight-paragraph episode 18, it has the highest word-to-paragraph ratio in Ulysses; so if by word count episode 14 belongs with episodes 16, 17, and 18, by paragraph count it is right next to episode 3.

Kenner was elated by the paragraphs in episode 14. He identified the 11+40+11 combination, but in all the excitement, even he miscounted—or at least pretended to for the sake of numerical symmetry.⁴⁷ If you add in the indented song, it’s sixty-three paragraphs (which also turns out to be the number of headlines in episode 7), and this matters because episode 14 would then break down into the three incantations at the beginning (“Deshil Holles Eamus”/“Send us”/ “Hoopsa!”) followed by sixty others mapping out six centuries of the English language, nine months of embryological development, and, according to one critic, tens of thousands of years of geologic time.⁴⁸ Prompted by Joyce’s own description in a letter, critics have focused largely on the many prose styles that get dispersed throughout. More specifically, they have treated episode 14 as a literary-historical crime scene, using Joyce’s own notebooks and sources to identify the fingerprints.⁴⁹

As fascinating as all the archival forensics can be, they’ve been ignoring the number and function of the paragraphs. The sum of sixty may evoke the mechanical clock of episode 3, but instead of also measuring the time of the mind, the scale has been radically enlarged in order to accommodate passing decades and centuries as strung together by an omniscient narrator. Time gets marked by changes in style, the thirty-one or thirty-two parodies of various authors and works corresponding with different historical periods (figure 1.3).⁵⁰ There is never any explicit indication in the episode that a new style has arrived, but that’s not necessary. The paragraph is there doing what it always does: providing line breaks and indents. Instead of marking changes in speaker or topic, the paragraph is there to clear the page for the arrival of a new style. So if each one functions like a time capsule, the breaks are where the travel from one decade or century to the other occurs.

Critics have long known about the stringing together of prose styles. What’s been missing from the discussion is the function of the paragraph (figure 1.4). Without the paragraph, episode 14 doesn’t work. New styles don’t just arrive anywhere at any time. They depend upon the line break and indent, so even in those instances when Joyce mixes different authors or even different chronologies, a new period gets marked by the arrival of a new paragraph. In modern terms, the paragraph is a place for transition, but that wasn’t always the case. In Middle English and Anglo-Saxon, it was an opportunity to break and begin anew, and Joyce would have learned about this from the History of English Prose Rhythm, a resource he borrowed from heavily. Though prose style is the main focus, George Saintsbury allows himself one valuable aside on the topic: “Of the paragraph, as such, it may be doubted whether any Anglo-Saxon or Middle English writer had much notion from the purely rhythmical-stylistic point of view. He had done with one subject and took to another: that was all.”⁵¹ Line break instead of rhythm and style: you can’t have it all.