Posted on 22 July, 2007 in category: Circles Disturbed Contributor Interviews, Interviews

The following interview between David Herman and Barry Mazur, and with a few comments by Petros Dellaportas co-founder of THALES+FRIENDS, was conducted on July 22, 2007.  The interview as it appears here was transcribed and edited only slightly for clarity.

Mazur: In your essay you give a panoptic view of the field of narratology, and of the various models that are used, which is extremely clarifying. You offer a view of the goals and meta-goals of narratology and you seem to be taking a very mathematical approach to your field.

Herman: I think I was trying to envision what concept might allow for maximum overlap (or at least an optimal means for comparison) between the domains of mathematics and narrative theory. Models and modeling processes struck me as being ideally situated at the intersection of these two domains. Hence I started to work backward, and review what kind of research on models had been done in fields that are at least allied with mathematics. Not having enough background in mathematics to know how mathematicians actually use models, I had to rely on the accounts of models outlined by philosophers of mathematics, theorists of applied mathematics, and so on. In composing the paper, I drew on that research as well as my own “native” knowledge of the models that circulate within the field of narrative inquiry.

Mazur: You refer to Max Black’s taxonomy of models.

Herman: In connection with Black’s taxonomy the notion of meta-goal that you mentioned before is quite pertinent. Or, to use an idea I develop in my paper, Black’s taxonomy has significant implications for metanarratology–that is, the study of what kind of theory the theory of narrative purports to be, or might aspire to be. If you look at models in the field and then try to place them somewhere within Black’s taxonomy, it seems to me that many of them fall in the category of analogue models. That is, they’re structural approximations of some sort: they try to mirror the
structure of stories, the web of relationships among elements of narrative. In working on my paper, I began to consider what it would take for narrative analysts to move from such analogue models to theoretical models, as defined in Black’s taxonomy. For Black, theoretical models, beyond functioning as structural approximations, have to explain or predict in some sense–in the way the computer models used by meteorologists predict
the weather, with more or less success. My impression was that in mathematics–and correct me if I’m wrong–there are such things as theoretical models, models that explain and predict and don’t just mirror. Accordingly, to try to figure out stepping stones that might lead narrative theorists from analogue models to theoretical models, I reviewed some implicit criteria that practitioners in the field seem to be using to evaluate existing
models. In exploring why scholars of story tend to favor some models over others, my larger goal is to work toward an account of why (and how) narrative theorists make such choices, using this sort of metanarratological inquiry to lay groundwork for new modeling practices in the field itself.
To be sure, there are already areas within the field that, leveraging empirical research, seek to explain and not just approximate the structures of narrative. Cognitive psychologists such as Richard Gerrig, for example, use empirical methods to investigate correlations between specific textual features and readers’ responses to those features, as measured by reading times. This kind of work might provide a pathway from analogue to
theoretical models. But in other parts of the field, the main focus of research does seem to be more about attempting to replicate through various kinds of models what a narrative is, what it looks like.

Mazur: In writing and thinking about narrative, why formalize at all?

Herman: I think that the formalizing enterprise or mission has been part of the field from its inception. Narratology, at a fundamental level, really is about narrative structures and the attempt to capture them formally. And here is where an understanding of formalism and formal models would help me–if I had a better sense of the range of uses of models in mathematics and in other, neighboring fields.

Mazur: Of course in mathematics one often develops a language to deal with some kind of formal structure. But usually one doesn’t do that unless one hopes that the structure will pave its own way–or pay its own way.

Herman: Otherwise one would just stick with the lower-level language, if I understand correctly?

Mazur: Yes. In mathematics I think there is a slight negative view of “museums.” Any model you build, if you simply put it in a case after it is built, is not paying its own way. Models should be used somehow. And your models are surely being used!

Herman: Yes. I think there’s something about narrative that requires distillation—for example, distillation from natural language texts to arrive at some sense of their underlying structure as stories. You can capture aspects of narrative through that sort of structural analysis that you couldn’t otherwise see. Analogously, if you parse sentences using linguistic models, you can see constituents and relationships among constituents
that would remain invisible without the help of those models. Indeed, the power of linguistic models to reveal such structural relationships is a main reason why linguistics came to be viewed as a “pilot-science” by the early narratologists; it inspired those theorists to assume that by developing models of narrative structure, you can capture aspects of stories that aren’t necessarily evident from their surface forms. Take, for example, Genette’s model of temporality and his account of the temporal relationships between the story plane (or what happened) and the discourse
plane (or how what happened is presented). Genette formulates exact ways of discussing order, duration and frequency, elaborating a systematic taxonomy. With respect to order, his model maps out what counts as chronological narration and what count as departures from it, such as flashforwards and flashbacks. Duration, in the model, can be defined as the ratio between how long it takes for an event to unfold on the story plane and how much text is devoted to its recounting on the discourse plane, with pauses, scenes, summaries, and ellipses corresponding to different narrative speeds. And frequency is the ratio between how many times something happens and how many times it’s recounted. Genette’s work is thus an example of a model that successfully captures and organizes a system of narrative possibilities–a system that you couldn’t easily envision without the model. You can look at individual narratives and see how they instantiate parts of that system.

Furthermore, by giving due attention to the features captured by the model I think you can avoid certain problems that enter into other traditions of narrative analysis. Are you familiar with William Labov and his work on conversational storytelling from a sociolinguistic pespective? Labov is particularly interested in what he terms narrative evaluation, his claim being that every storyteller has to underscore the point of his or her story, signal its reason for being told in a particular context of telling, to ward off the dreaded question “So what?” If you’re telling a story and somebody
interposes this withering rejoinder (to use Labov’s phrase), then you know you’ve failed. By evaluating the events being presented in the narrative, you can highlight the narrative’s point and thus avoid the “So what?” question. In turn, one of the strategies that Labov uses to describe evaluation is to posit what he characterizes as a standard or default syntax for narrative, whose basic unit is the “narrative clause.” For Labov, narrative clauses are such that they cannot be transposed without changing the underlying semantic interpretation of the story; furthermore, he argues that these default narrative clauses are very simple, marked chiefly (in English-language storytelling) by past tense verbs in the indicative mood. The sequence of narrative clauses featuring these verbs iconically matches the order of the events they report, and the “baseline” clauses themselves are devoid of modal auxiliaries, negatives and hypotheticals, and other complexities of structure.

But are such complexities of structure always evaluative in their function or force? In contrast with Labov’s model, Genette’s model suggests that any of the structural elements encompassed by the narrative system may or may not be deployed in a marked or event-evaluating way–from “iterative narration” via constructions involving the modal auxiliary would (we would go to the ocean often that summer) to time-bending flashbacks and flashforwards. Conversely, Genette’s model captures a resource for evaluation not registered in Labov’s account: namely, the way fluctuations in narrative speed can be used to mark off salient actions and events from background circumstances and happenings, as when a narrative shifts from a summarizing to a scenic rate of presentation.

Mazur: Is Labov’s account an empirical summary, though? Possibly Labov is focusing on the story-tellers he’s studied, rather than the general range of story-writers.

Herman: Yes that’s true. That’s one way to justify the model Labov has produced. He models precisely what he found in the specific corpus of narratives he looked at. In contrast, Genette’s work focuses on literary narratives, where other constraints on communication are in place. Written narrative allows for more complex structures because readers can flip back through the text, reread the same sentence multiple times, look ahead, etc. So there’s a way in which, you’re right, the communicative situation might explain the contrast between Labov’s model and Genette’s.
But actually there have been follow-up studies by researchers also doing empirical work on face-to-face narration. They argue that Labov was looking at one subtype of narratives produced in interactional environments where an interviewer prompts somebody to tell a story and then stands back as that person tells his or her narrative in a relatively monologic way. But there are many, many other types of interactional environments involving co-narration, intense competition for the floor, he-said-she-said gossip, and so forth, and so other analysts, seeking to diversify the corpus
of stories on which frameworks for inquiry might be based, have developed models that attempt to capture the full range of storytelling situations. There’s a tremendous book–it’s the best book on storytelling in face-to-face interaction that I’ve ever come across–called Living Narrative, by Elinor Ochs and Lisa Capps (Harvard University Press, 2001). Ochs and Capps develop the notion of narrative dimensions, including linearity, tellership, tellability, embeddedness and moral stance. Any storytelling act will be located at different points along these five dimensions, with Ochs and Capps thus providing a highly nuanced picture of the different systematic possibilities available to people using narrative in face-to-face discourse. So theirs would be another model that we could talk about, alongside Genette’s and Labov’s.

Mazur: There is, it seems, a link between some mathematical developments and what goes under the heading of structuralism in anthropology, and–I would imagine–in structuralist aspects of literary theory. Can you tell me more about that?

Herman: You know, this link has certainly been prominent in the history of the field. And actually I would like to hear more about Bourbaki from you because I only really started to read about the group while doing research for my paper, though I’d known vaguely about Roman Jakobson’s interactions with some of the group’s members. In general, structuralism was an approach to literary and cultural analysis, especially prominent
in the 1960s and 1970s, that used linguistics as a pilot-science (in the sense mentioned previously) to study diverse forms of cultural expression as rule-governed signifying practices or “languages” in their own right–with narratology being an outgrowth of this general approach. Before the French structuralists, however, the Russian Formalists had already begun to develop important tools for narrative analysis back in the early part of
the twentieth century. Commentators like Viktor Shklovskii, Boris Tomashevskii, and Vladimir Propp were affiliated with that group. These scholars developed for example the fundamental distinction between story (fabula) and discourse (sjuzhet), which made its way to France partly because of Tzvetan Todorov, whose knowledge of Slavic languages helped bridge the Russian Formalist and Francophone structuralist traditions. The high point of structuralism was in the mid to late sixties, and the term narratology itself was coined by Todorov in his 1969 book Grammar of the Decameron, which presented a structuralist analysis of Bocaccio’s text. Like other domains of structuralist theory, structuralist narratology was based on a Saussurian understanding of language, which divides language into langue or system and parole or message (or utterance). Further, the structuralist narratologists, following Saussure, wanted to background individual narrative messages (or texts) in favor of the larger narrative
system, or narrative langue. For Saussure, the message or individual utterance is too contingent, too dependent on factors that cannot be accounted for in linguistic terms. The structuralist narratologists likewise sought to focus on the system that makes possible the production and understanding of individual narratives. They thus argued that narratology should not be viewed as a handmaiden for interpretation; its goal is not to interpret
individual narrative texts but rather reveal their underlying structure–just as the linguist does with sentences.

This broadly Saussurean approach led in turn to an uncoupling of structural analysis from literary interpretation, although the narratologists did recognize that attention to these structural possibilities could enable interpretation. That is, if you could set out the full system of narrational possibilities it might allow you to say something interesting about why you have the interpretation that you do of, say, Proust’s In Search of Lost Time. With Proust, a key issue is his tendency to use iterative narration (a type of frequency that involves telling once what happened more than once), so in this case the narratological categories do in fact allow for a more precise analysis of the temporal structures of Proust’s novel.

In the decades since they were first formulated, however, structuralist approaches to narrative have come under attack from a number of directions. One line of criticism has been that narrative theorists really do need to engage with issues of interpretation–on pain of having to face the “So what?” question in their own right. Another critique concerns the language theory that the structuralists derived from Saussure’s ideas, which aren’t necessarily the best for capturing how language works in situated communicative contexts. When it comes to studying aspects of language in use, other theoretical frameworks are arguably more powerful—frameworks such as pragmatics and discourse analysis, which focus on how using certain language designs in certain kinds of contexts licenses inferences about utterance meanings, as discussed in the work of Grice, for example. These traditions of linguistic inquiry are particularly important for understanding extended stretches of text or talk–that is, discourse. So the structuralists
were in some cases working with less-than-ideal linguistic tools when they investigated how narratives function as structures. In some of my own work I’ve tried to retain the basic structuralist insight that language theory provides invaluable resources for narrative study, while drawing on different, more recent linguistic concepts and methods in an effort to flesh out that original insight.

Mazur: Are the formal models of narratology analogous to the formal models of music in musicology? Using formal models of music enables you to talk about and to decide what key you’re in. That then manages to structure your experience of the music for it determines what the dominant is, what the subdominant is; it tells you how certain harmonic progressions will end up, or at least suggests such endings, giving you expectations and anticipations. This kind of judgment helps the listener to hear the intentions of the composer and helps the composer to organize the piece, helps the
performer to perform it so you’ve got the composer, the listener and the performer, all in a kind of loop.

Herman: That’s a great point and I’ve thought about some analogues in narrative theory. Generic codes would be clear parallel here, since genres too set up expectations and anticipations, allowing interpreters to channel and delimit their inferential activities as they engage with a particular text. Or take musical compositions involving both horizontal and vertical dimensions, or melody and harmony. Analogously, some of the models of plot feature not only the linear unfolding of events over time, but also interrupted or truncated action lines that are caused by conflicts among characters–conflicts that cause some of the characters to refrain from performing actions or attempting to pursue their goals. So then you have a range of ghostly, unrealized action lines paralleling or shadowing the main plot line–which would mirror compositional structures involving melodic as well as harmonic relationships. Another parallel, in this same connection, might be afforded by multiplot narratives that follow several plot-lines, involving different groups of characters–as in the Victorian novel in the British tradition. As for modulations of key, perhaps one could talk about adaptions that shift narrative registers, as when a story about serious events is renarrated humorously? I think the disanalogy here would be that–with the exception of work on storytelling in face-to-face interaction and some of the emergent work on computational narratology–narrative theory has largely focused on the reception side of things rather than on story generation or production. True, in the tradition of sociolinguistic analysis, researchers working in the area known as Conversation Analysis have developed a method for analyzing ways of producing discourse. Conversation Analysts assume that people signal their understandings of situations by how they talk; hence these scholars explore how interlocutors create social situations in part by producing particular kinds of utterances–utterances that index a certain understanding of what is going on and thereby help bring about, in a self-fulfilling way, the type of situation that corresponds to the interlocutors’ own sense of what’s happening. This research has examined how storytelling works and, more specifically, how building narratively organized sequences in discourse requires–and involves the socially coordinated display of–certain attitudinal stances and strategic maneuvers. So this approach does get at the design element of narrative.

But in literary narratology the focus has been more on processes of reception. That said, however, there is a strand of research on literary narratives that is rooted in rhetoric, which of course studies how communicative situations involve language producers as well as language interpreters. This tradition was pioneered by Wayne Booth, in his classic 1961 study The Rhetoric of Fiction.

Mazur: Isn’t there a moral thrust to Booth’s work?

Herman: Arguably, there is. Two literary critics, Wimsatt and Beardsley, wrote an essay in 1946 called “The intentional Fallacy,” in which they objected to methods of interpretation that invoked the intentions of the author. Partly in response to this line of argument, I would suggest, Wayne Booth developed the concept of the implied author. One of Booth’s concerns was how readers might be affected by texts that didn’t overtly censure narrators or characters with suspect values. If you define the implied author as a kind of value system and explore whether a narrator or a character is favored or disfavored in the terms afforded by that system, then it’s a natural step to extend the focus of concern to how readers are impacted by this process of negotiating the value system (or systems) bound up with a given text.

For what it’s worth, in my contributions to a forthcoming co-authored book, titled Practicing Narrative Theory: Four Perspectives in Conversation (co-authored by David Herman, James Phelan, Peter J. Rabinowitz, Brian Richardson, and Robyn Warhol, and in press at the Ohio State University Press), I’ve developed a critique of approaches based on the concept of the implied author. (My co-authors, I should stress, defend and deploy the idea of the implied author.) I’ve also proposed there a way to explore issues of value that, drawing on the concept of norms, contrasts with Booth’s.

Mazur: Modern American authors are sometimes not only not implied but are rather quite explicit and try to preserve the literary imagination from too four-square an insertion of ethical strictures; for example, Philip Roth once said: “in my imagination I betray everybody.” Value judgments regarding narrators must become particularly complex when there are layers of different narrators involved in the telling. Genette, for example,
discusses the different narrator-strands in the Odyssey.

Herman: Yes, that’s right, and in fact theorists have developed a variety of models to account for this phenomenon. One such model is basically a vocabulary of levels deriving from Genette’s work. In this model you have a sort of matrix scenario, or baseline narrative reality, and then embedded scenarios–so that as you descend within the frame structure you arrive at a more and more distant remove from the matrix frame, so to speak. This is where the term diegetic comes into play. The diegesis or the diegetic level is the “primary” level or frame evoked by a narrative text, and a hypodiegetic level can be created if within the main narrative level a character starts to narrate. So in Chaucer’s Canterbury Tales, for example, there is the baseline reality of the pilgrimage in the context of which the characters tell stories to one another; those characters are thus intradiegetic narrators who tell hypodiegetic narratives. This analytic machinery is an attempt to capture how texts can have different narrator-strands, as you put it, and it provides a basis for exploring how different kinds of value judgments might be more or less pertinent depending on which strand you’re focusing on.

Mazur: Is it possible that narratological theory could be used as a schema for the language in which we tell anecdotes about our daily life?

Herman: There have indeed been attempts to use narratology for this purpose–for example, in one of the first systematic contributions to this field, a 1966 essay by Roland Barthes titled “Introduction to the Structural Analysis of Narratives.” Barthes argued that there are two kinds of fundamental plot-units: kernel events (or “nuclei”) and the satellites (or “catalyzers”) that expand upon them. In the terms afforded by Barthes’ model, one can say that to paraphrase a story you would at least have to capture the kernel events–whereas how many of the satellites you bring along with them is negotiable and will vary according to the needs of the paraphrase. On the production side of things one could deploy this same model to talk about how a person narratively configures his or her past experiences. In telling a story about some formative event, I know that I need to include at least some kernel events. Further, an implicit principle of relevance then allows me to determine how many and which satellites to include. And the metric of relevance will change with the context–depending on your audience, what kind of narrative you want to craft, and so forth. Is it for a family scrapbook? Is it for an e-mail message?

This is just one (small) example of how ideas from narrative theory might throw light on the structures and dynamics of everyday storytelling. Again, I’d recommend Ochs and Capps’ Living Narrative for its brilliant discussion of a whole range of relevant issues.

Dellaportas: Can I ask you something related? In a model focusing on music, the temporal dimension is key. Thus, when you construct a model of music you would expect a certain combinatorial argument there. I’ve seen a lot of studies in literature that count how many times certain things happen etc. Likewise you would expect narratological models to take into account this temporal dimension. For example, after highly dramatic moments you would expect something calmer. Do narratological theories take this sort of temporal rhythm into account?

Herman: Some of these issues have been explored under the umbrella category of plot—that is, the study of plot structure and the overall trajectory of a narrative over time. Thus the German theorist Gustav Freytag developed the so-called Freytag’s triangle to account for the structure of plots; the triangle suggests a movement from the introduction to the climax and thence to the denouement. Subsequently, Meir Sternberg developed a different account of exposition and temporal ordering, which suggests that the exposition of background material can be differently distributed in a given narrative. It can be concentrated in the beginning or else dispersed in various places throughout the text. Sternberg also draws a distinction between exposition and scene and argues that each work establishes its own “scenic norm,” which is a pace of narration that suggests you are no longer in the domain of exposition but instead in a scene-like structure which is important and therefore catches your attention. So Sternberg is interested in the play between exposition as background, which narratives typically provide in a relatively rapid way, and scenes, which the text presents in a comparatively slower and more detailed manner. Work of this kind does try to take temporal rhythm into account.

Mazur: I’m guessing that narratology is not an attempt to model our internal thinking. Is this right?

Herman: Some of the recent work in narrative theory does try to connect up three domains: (1) humans’ underlying cognitive abilities; (2) linguistic and more broadly semiotic reflexes of those abilities; and (3) aspects of narrative structure that bear on both (1) and (2). Some of the research on narrative perspective (or, to use the technical term, “focalization”) that I discuss in my paper seeks to explore links among these three
domains. Further, another tradition of inquiry in the field draws from the interdiscipline of narrative psychology, a rich, integrative approach that takes inspiration from some of Jerome Bruner’s foundational work. This approach combines ideas from the philosophy of mind, social psychology, and discourse analysis and tries to move away from a notion of the mind as internal and instead tries to embed the mind, or mental processes, in
discourse–and in narrative as key type of discourse. From this perspective, the mind is less something located inside the head than a structure spread out among participants in narrative, the story itself, and the setting in which the narrative occurs. Hence we could be talking about something cognitive that would not however be internal–or at least not wholly internal. In current work I’m exploring some of the implications of this approach when it comes to studying the nexus of narrative and mind.

Mazur: Is there a piece of literature that makes you feel “if I didn’t have this equipment, the narratological models, I would be less of a reader of that piece”?

Herman: I’ve certainly had moments in which I felt that I had to pull every narratological trick out of the hat to try to figure out why I was affected by a text in the way I was. Here I think that one again works in a backward manner. I mean that you get struck by something in a text and then you try to figure out why that aspect of the text had such an impact on you. You ask yourself: what aspect of the narrative’s structure has affected me in this way, and how can I best describe it? This is where narratological models become helpful—and fruitful. For example, many postmodern narratives involve a structure that Genette identified as “metalepsis.” Metaleptic narratives are marked by a conflation of diegetic levels. At issue is a sort of M. C. Escher-like phenomenon, where a hand comes out of the page and draws the other hand, so that you’re not sure what’s the actual world and what’s the imagined or create world. That would be a metaleptic pattern: the embedded layer swallows the embedding layer endlessly in a kind of looping structure. There are stories with this same structure, such as Julio Cortázar’s “A Continuity of Parks,” where a character reads a story in which he, that same character, is about to be murdered. In this context, knowing about the theoretical frameworks allows you to appreciate for
one thing the subtle complexity of the writer’s design. It also allows you to assign a structural description to the text that provides an account or basis for your reaction. You can say, “that’s why I found it so strange!” You can do this intuitively, of course, but with
this vocabulary you can develop a fuller account of the text’s defamiliarizing effect, and also think comparatively about how that effect plays out in other, more or less closely analogous texts.

Mazur: We’re TV viewers and film watchers and I don’t know much about this but I do know that the manner in which scenes are cut changes drastically from decade to decade. There is nowadays a quicker tempo; a lot more is assumed now because we’re very practiced viewers of film, so we can manage when scenes quickly cut into other scenes. And yet there are certain moments, certain cuts that, apparently, are never done. There are still restrictions but there’s an evolution of the class of allowed cuts; I wouldn’t call it progress, but I want to call it a trajectory of change based on viewers’ cinematic experiences. Do you see that in the representation of time in literature and do you see a way of integrating a dynamic component into narratological theory? If you had this dynamic component, you might be able to predict, like the linguist can vis-à-vis changes in the vowel-system of a language, future changes in the narrative system when it comes to temporal structures.

Herman: That’s a great question. In fact, James Morrison, in his contribution to a volume titled Teaching Narrative Theory, writes about the shorter scene durations in modern films. Shots used to be something like an average of nine seconds long in classic Hollywood films, like Orson Welles, but now they are something like three seconds long.

Mazur: There’s got to be a literary equivalent to this and surely narratological theory is going to be an absolutely crucial tool for understanding these dynamics.

Herman: There have been efforts to build in a dynamic component of this kind—not so much in the area of temporality, but rather in the area of narration, and more specifically in the presentation of characters’ consciousness. The Austrian narrative theorist F. K. Stranzel developed an approach of this kind. He identified three main narrative situations: the authorial narrative situation, which involves third-person narration and a distanced view of events; the first-person narrative situation; and the figural narrative situation, which features third-person narration but a filtering of the action through a character’s consciousness—as you get with Virginia Woolf, Henry James, James Joyce, or Franz Kafka. And Stanzel’s argument is that although the figural possibility was always in play more or less throughout literary history, it became concentrated only at the end of the 19th century into the early 20th century. Indeed, Stanzel’s approach has led to the proposal for a diachronic narratology, which would look at how narrative structures might have changed over time—how different structures or techniques were used to produce a given effect, and how the same technique may have acquired different functions over time. I’m not sure whether this approach has predictive value, but at the very least it has historical scope.

I had the privilege of editing a recently published volume that explores directions for inquiry in this subfield of narrative research–that is, diachronic narratology. The volume is titled The Emergence of Mind: Representations of Consciousness in Narrative Discourse in English (University of Nebraska Press, 2011). It features contributions by specialists in different literary periods who use specific case studies to explore changing trends in the representation of fictional minds in English-language narratives. All in all, the volume covers the period stretching from around 700 to the present.

Mazur: It might be particularly useful to study the evolution of narrative structures in the work of a single author.

Herman: Joyce would be a good example.

Mazur: Do you use narratological models to study Joyce? And if so, what do you think about Frank O’Connor’s remark that Joyce’s “The Dead” is something like a novel, in contrast with Ulysses, which he thinks of as a short story?

Herman: Yes, I’ve looked at a number of Joyce’s texts from a narratological perspective–they are so rich and multidimensional! Regarding O’Connor’s comment, my understanding is that Joyce did originally intend to write a short story called “Ulysses,” but that his subject refused to be limited or constrained in that way. But beyond these specific considerations, I would reply to O’Connor by asking a more general question: how do you define a short story (in principle)? What constitutes the dividing line between short story and novella, for example? I don’t know if there’s an exact divide, but it’s an interesting question. Is it size? Is it complexity?

In any case, there are lots of interesting narratological questions to explore when it comes to Joyce, and different strategies for posing those questions. In my chapter on “The Dead” for The Cambridge Companion to Narrative, for example, I draw on several traditions of research to examine how Joyce goes about representing his characters’ minds.

Mazur: Do you discuss the maid? What’s the maid’s name?

Herman: Lily.

Mazur: Her speech is about the dead, and about how men nowadays are all palaver and what they can get out of you. Do you make out the resonance of the word cadaver in palaver?

Herman: That’s a great point! Indeed, the opening scene with Lily exemplifies the structure which has been called free indirect discourse, which happens when a character’s voice (or mind style) inflects a narrator’s voice. Another interesting aspect of Dubliners is that it is a short story collection but the stories are linked. Some of the characters, situations, and events seem to cross over the boundaries between individual stories,
which is an interesting narrative structure in its own right. In “Two Gallants,” one character’s last name is spelled “Corley” but it’s pronounced “Whorely,” and sure enough he deceives a housemaid (like Lily) into giving him her whole month’s salary.

This incident resonates with Lily’s comment, in “The Dead,” about how men try to take advantage of women. There’s a finely wrought language pattern here, one typically associated with poetry rather than prose.

Mazur: David, the sweep of your analysis covers panoptically the grand structure of narrative as well as the syllable-by-syllable music of imaginative literature. I think you wrote a wonderful essay. Thank you!

Herman: Thank you, Barry, for your excellent questions!