The chart below depicts the length of paragraphs in Joseph Conrad’s Heart of Darkness. The paragraphs are listed in order, first to last, from left to right. The length of a bar is proportional to the number of words in the respective paragraph.
There is an important issue at stake: Does that chart depict a phenomenon that is real, not only in the sense that it represents a set of valid measurements, but in the deeper sense that those measurements represent some intrinsic aspect of the causal process that produced the text. For the most part I leave that issue to the discussion I had with Claude 3.5 Sonnet.
The purpose of this introduction is to indicate why the issue arises.
The problem of literary form
Despite the fact that form and formalism are important concepts in academic literary criticism, the careful description of form is not central to current practice. This is not the place to discuss this issue in any detail. I have done that elsewhere on New Savanna, I have published a long article on form (Literary Morphology: Nine Propositions in a Naturalist Theory of Form), and have written a number of working papers on description.
Form is an aspect of the physical substance of the work of art. That is quite obvious in the case of drawing, painting, and sculpture. That is also true in the case of music, where we are interested in the arrangement of sounds in time, and dance, the arrangement of gestures and movement. Literature is like that as well; it is a string of words, one after the other after the other after the other. It follows, then, that the description of literary form involves analyzing how words are placed in a string.
And there is some of that in literary criticism. In dealing with poetry one may learn how to count feet in a line and work out the rhyme scheme, if there is any. But this is often treated as decorative and unrelated to meaning in any deep way, and it’s meaning that critics are after. In the study of narrative the Russian Formalists made a distinction between story (fabula) and plot (syuzhet), where the former refers the events as they occur in order while the latter is the order in which events are introduced into the narrative. In the extreme case of Tristram Shandy, for example, plot and story are quite different. This distinction, and more, is quite important in narratology, but narratology is a relatively minor approach to literary criticism and one that is more important in Europe than the United States.
As an example of this situation, here’s a few lines from an essay by Frances Ferguson and John Brenkwood introducing papers from the 2013 English Institute on form:
A second irony is that the recently renewed interest in questions of literary form has proved quite amorphous. Perhaps, though, that has been the predicament and vitality of the topic all along. [...] There is also a conflict, cutting across these competing methods, between form as a feature of literary works and form as constitutive of literary works. The New critics are often the benchmark of formalism in American discussions, but they did very little to illuminate literary forms compared to the Russian Formalists or, say, Lévi-Strauss and Jakobson’s classic essay on Baudelaire’s “Les chats.” [...] And yet even the surest markers of literary forms fail to define form when it comes to actual works. The form of the sonnet, for example, is readily defined by the number of lines and the stanza organization, but does that account for a particular sonnet’s form any more than a rectangle accounts for a painting’s form? Vertical for portraits, horizontal for landscapes! And, finally, is formalism itself based on the idea that literary works are purely form, or on the idea that the vocation of literary criticism lies in formalization, that is, in its capacity to create categories at a level of abstraction applicable to the widest variety of literary phenomena?
The upshot of this conundrum is that literary form is more an object of wonder and a source of methodological argument than the basis of practical criticism. It’s important, but we don’t know much about it.
Thus, that chart is something of an embarrassment to literary critics, or would be if they saw it. On the one hand, it represents and objective fact about Heart of Darkness. When you count the words in each paragraph and graph the result, that chart is what you get. [I used the text provided by Project Gutenberg.] Once you’ve seen it, you can’t argue it out of existence.
Now, I’ve been in this business a long time. While I’ve discussed that chart and its implications extensively in my working paper on Heart of Darkness, I’ve not bothered to seek formal publication. Why not? Opportunity cost. I don’t know of any venue that would be interested and I don’t want to waste my time jumping through hoops making that work acceptable to some journal only to wait six months to a year or more to have it turned down (though I’m seriously considering taking a shot at it in the new year; we’ll see).
However, when I discovered that structure I did send at note to Mark Liberman, who runs, Language Log, and group log devoted to linguistics. He was kind enough to post a link and we had a lively discussion, Markov’s Heart of Darkness. The discussion started with the issue of whether or not paragraph length followed a power law as that’s the question I presented to Mark; it does not. But one Lucy Kemnitzer brought up the issue of order during the discussion and I linked to a post in which I had the chart I posted above. If you look through that discussion, you’ll find comments by someone identifying themselves as “JL” and noting that they are a novelist. They’re quite dismissive of that kind of description and the analytical work it supports:
The question is whether you're adding anything at all to anyone's understanding of the book, or whether, instead, you're indulging in puffery and pseudoscience, bringing in spurious mathematical models and childish anthropological "theories" to gaudy up a banal and perfectly obvious account of a very well known phenomenon. (Look! I've discovered that the ratio of vowels to consonants in 'The Waste Land' is exactly equivalent to Planck's Constant! Which is used to explain black holes! See? It really is a depressing poem…)
That’s the kind of reaction I expect of well-trained humanists.
Ask Claude
And so I turned to Claude 3.5 Sonata. What would they think of that distribution of paragraph lengths?
I’d already loaded the entire text of Heart of Darkness into Claude and had some discussion about it (I’ve already posted some of that discussion). It was a simple matter to reactive that discussion, upload the chart, and initiate a discussion. The discussion went well, at least from my point of view. Claude even pointed out some things that I hadn’t noticed.
But after all the (standard) literary criticism it surely must have ingested during training, why didn’t it react with, if not outrage, certainly skepticism? Why wasn’t it aligned, as the kids say, with the conventional wisdom on this matter? Context, I figured, context. This interaction was loaded with Conrad’s text and extensive earlier discussion. Standard literary criticism was nowhere to be found in the immediate context.
“I wonder,” thought I to myself, “if I could create a context in which Claude would register a conventional reaction to that chart?” No sooner had I registered that thought than I figured out how to do it. I had previously loaded Claude with the Hamlet chapter from Harold Bloom’s, Shakespeare: The Invention of the Human. I’d asked Claude to summarize the chapter and then we had a little discussion.
So, I reactivated that discussion, uploaded the chart, and BOOM! Claude reacted with the expected skepticism, albeit a rather mild expression of it:
Based on Bloom's critical approach, he would likely be skeptical of this kind of quantitative or structural analysis of literature. He would probably view this chart as an example of what he saw as the problematic tendency in modern literary criticism to focus on technical or formal aspects rather than aesthetic and imaginative power
The discussion went on for a bit, took an interesting turn, at which point I decided to introduced Susan Sontag’s 1964 essay, “Against Interpretation,” into the discussion. Why? Because she argued that “criticism needs to focus less on trying to extract hidden meanings from art and more on describing the direct sensory and formal experience of engaging with it” (Claude’s words). We had an interesting discussion. For what it’s worth, Claude agrees with my view that literary criticism needs to pay more attention to the analysis and description of literary form.
An exercise for the reader: What does Claude’s ability to take different and contradictory positions on the same issue – in this case, the value of that chart for analyzing Heart of Darkness – suggest about the much-discussed problem of AI alignment?
* * * * *
First, we have my discussion with Claude about Heart of Darkness. Once that’s finished, I introduce the Harold Bloom discussion, which leads to Susan Sontag. As always my prompts are in bold face while Claude’s responses are plain.
I’ve uploaded a bar graph. Each bar represents a single paragraph in Heart of Darkness. They are arranged in order, with the first paragraph on the left edge and the last paragraph on the right. Do you have any observations about paragraph length as revealed in this chart?
Looking at this visualization of paragraph lengths in "Heart of Darkness," several key observations emerge:
Most paragraphs in the novel are relatively short, with many falling under 200 words. This creates a baseline rhythm for the narrative. However, there are significant spikes where paragraphs become much longer, particularly in the middle section of the text.
The most striking feature is a dramatic peak of approximately 1,400 words around paragraph 109-110. This extensive paragraph appears to be an outlier, being significantly longer than those around it. There are also several other notable peaks reaching between 600-1,000 words.
This variation in paragraph length suggests a deliberate structural choice by Conrad. The longer paragraphs likely correspond to key moments of exposition or critical narrative developments, while the shorter paragraphs maintain the story's forward momentum and may handle more immediate action or dialogue.
The pattern also shows that the novel ends with a series of consistently shorter paragraphs, perhaps indicating a more rapid narrative pace as the story reaches its conclusion.
This visualization effectively reveals Conrad's technical approach to pacing and structure in the novel, showing how he varies paragraph length to create rhythm and emphasis throughout the text.
Would you be interested in identifying what narrative content appears in those particularly long paragraphs? That might help us understand why Conrad chose to expand certain moments in the story.
You’d think that variation was deliberate. But we have no evidence that Conrad was conscious of this. He just did it. It’s how he wrote.
You make an excellent point. I fell into a common critical trap of assuming intentionality where we have no evidence for it. This pattern of paragraph length variation emerged naturally from Conrad's writing process - it's simply his natural rhythm as a writer.
This observation connects to a broader point about literary analysis. While we can usefully identify patterns and structures in texts, we should be cautious about assuming these were consciously planned by the author. Sometimes the most powerful effects in literature arise organically from how a writer naturally approaches their craft.
In Conrad's case, this varying paragraph length creates compelling effects in "Heart of Darkness," but we don't need to attribute this to conscious design. It's simply part of how he told stories, his natural way of allowing some moments to expand while keeping others concise.
This is a useful reminder about the difference between identifying patterns in a text and making claims about authorial intention.
That single longest paragraph, I believe it’s 1502 words long, is that nexus paragraph we’ve already discussed.
I’ve already had extensive discussion about this paragraph with Claude. Here's part of that discussion, though it doesn't contain Claude's remarks about the nexus paragraph. However, here’s my original blog post about that paragraph. And this post places that paragraph in the context of the whole text.
That's a fascinating discovery about the nexus paragraph! The fact that it’s not only structurally and thematically central to the novel, but also stands out quantitatively as the longest paragraph, adds another dimension to our understanding of its importance.
This paragraph accomplishes several crucial narrative tasks: it provides Kurtz's European background, introduces his report with the chilling "Exterminate all the brutes" line, and frames the death of the helmsman. The paragraph's exceptional length seems to naturally correspond to its function of bringing together multiple strands of the story.