Narratives give structure to our lives. We make sense of our daily observations and interactions with others through stories. Even our descriptions of seemingly minute events bear a resemblance to narrative structure, with beginnings, middles, and ends; the “frames” in Frame Semantics[i] are not so different from micro-narratives[ii]. I believe that the study of narrative can actually yield great insights into the ways in which humans understand the world.
I’ve always enjoyed reading, but I only recently developed an interest in narrative theory. Below is a very quick introduction to some of the general ideas, with particular emphasis on Vladimir Propp’s work; later posts will likely go into more detail, and connect this back to more testable theories in cognitive science and computational linguistics.
Narratology, the study of the structure of narrative, was officially born in the 20th century, though investigations into the nature of narrative date at least to Aristotle’s Poetics. It was in the 20th century, however, that literary theorists began thinking about how best to formalize a theory of narrative. Generally speaking, formalizing a theory of domain structure is the point at which research can begin to ask and answer difficult questions about the domain. In an essay on the structure of myth[iii], Claude Levi-Strauss pointed out that the problem of formalizing mythological structure is not unlike the problem faced by linguists a century before: how can the apparent arbitrariness of individual myths be reconciled with their fundamental universal similarities, as well as a sense of “timeless relevance”? Linguists solved this problem by breaking language down into structural units, such as “phonemes” (units of sound in a language) and “morphemes” (the smallest grammatical units in a language).
Levi-Strauss proposed that mythologists develop a comparable unit, which he alternately called a gross constituent unit or mytheme. He argued that mythemes are compositional in nature, and like other structures in language, are built from the lower-level structures; in this case, mythemes are constructed at the sentence level, in the descriptions of events and their participants. Levi-Strauss proposes a “paradigmatic” approach to the study of myth, meaning an approach focused on thematic or “latent content” embedded in the text. That’s about all I’ll say regarding Levi-Strauss in this post, though I highly recommend checking out his essay The Structural Study of Myth for an example of his structural analytic approach to the Oedipal myth.
Functions of the Dramatis Personae
Vladimir Propp, author of Morphology of the Folktale[iv], also advocated for a more rigorous and structure-focused study of narrative, though his view was “syntagmatic”, meaning it focused on the linear structure of events – e.g. how certain events lead to other events. Using Russian folktales as his set of corpora, Propp argued that all narratives can be described in more general patterns or abstractions (e.g. THE VILLAIN CAUSES HARM TO A MEMBER OF THE FAMILY), which are instantiated in more specific events in the actual tales (e.g. the “swan-geese” kidnapping a girl’s little brother).
Propp coins his own term for these patterns, calling them “narrative functions”, which are carried out by the various characters (dramatis personae) in the story. Two individual events in separate narratives can be described with the same narrative function if they “advance the story” in the same way. Ultimately, Propp argues that these narrative functions ought to be used as a framework for dissecting, classifying, and comparing folktales.
Prior to Propp’s work, most syntagmatic approaches in narratology suffered from two main issues (according to Propp):
- No “systemic description”: there was no well-defined universal terminology for folktales, such as what exists for mathematics and physics.
- Imposition of structure: classifications were usually imposed, rather than extracted, from the narrative.
Many of the traditional methods in categorizing folktales end up violating the “rules of division”; for example, the classic division of folktales claims that “all stories are either about animals, everyday life, or fantastical things”[v]. These strict classifications fall apart upon closer inspection; many tales involving animals are also about fantastical things, and many tales about everyday life involve animals. This classification also fails to identify the commonalities between tales, such as identical actions taken by different characters. Even one of the more well-known classification schemes, known as Aarne-Thompson tale types, falls into this trap[vi].
The most relevant precedent for Propp’s approach is the idea of motifs, put forth by the theorist Veselovskij. Motifs are indivisible narrative units, which combine to form narrative themes. Propp essentially agrees with this approach; his only criticism is that Veselovskij’s examples of motifs can actually be decomposed further into narrative elements[vii]; these elements are the subject of Propp’s inquiry.
Fundamentally, Propp’s idea of narrative functions is rooted in his claim that tales often attribute “identical actions to various personages”; many tales describe similar sequences of events with similar causal connections between those events, even if the specifics vary. Importantly, this doesn’t mean the same actions always perform the same function; a hero receiving 100 rubles from his father to purchase a “wise cat” is a distinct narrative function – and comes at a different point in the narrative – from a hero receiving 100 rubles as a reward for bravery. In Propp’s view, functions are the “building block” of narrative; he claims that the set of functions is finite, and that they always appear in the same order. That is, if two functions A and B both appear, they will always appear in the same order. Propp formalizes a typology for these functions[viii]:
- Summary of the purpose of a function, e.g. THE VILLAIN CAUSES HARM OR INJURY TO A MEMBER OF A FAMILY.
- A one-word definition, e.g. villainy.
- A conventional sign, e.g. A.
The other important point to note is that the ways in which these functions connect to each other – the structure of function sequences – has a particular logical structure. Certain functions entail a later function, or reference previous functions. For example, the function HERO RECOGNIZED BY HIS BRANDING references the earlier function HERO RECEIVES A BRANDING. Some of these functions can be composed into higher-level abstractions – which would perhaps be analogous to motifs – that represent consistent groupings of particular functions[ix]. For example, the sequence of villainy, dispatch, decision for counteraction, and departure from home compose to form the “narrative complication”.
Caveat 1: Propp describes many other insights regarding narrative structure and functions, but I’m not going to go into detail here. That said, I’ll likely discuss this again in a later post – or you can just read his book.
Caveat 2: If you’re of a suspicious mind, you’ll likely be skeptical of Propp’s claims. Note that his allegedly exhaustive survey was specifically on Russian folktales, and in fact, Propp explicitly mentions that he makes no claims as to the narrative functions present in:
- Folktales more generally
- Epic tales and legends
- Anecdotal tales
…And so on. That said, he does seem to think that the general notion of a narrative function can be applied to pretty much any domain of narrative, just as the structural units of language that motivated narratology can be used to describe phenomena in many different languages.
Why This Matters
I’ll address the significance of narratology in more detail in a later post, but here’s a quick summary of why this matters to me, and why I think it should matter to other people interested in language and the mind:
- If narratives can all be described with a similar structure, that might say something about human conceptual structure – the ways in which humans prefer to organize knowledge about events in the world. This would be cool.
- Artificial intelligence – and specifically, machines that understand language – is one of humanity’s grandest aspirations, with both immediately practical and loftier applications. A machine that could understand narratives – or, as in Roald Dahl’s The Great Automatic Grammatizator[x], generate narratives – would appear to possess an uncanny level of intelligence. This, apart from also being quite cool, could be of use to literary theorists, as well as those trying to make sense of all the text-based content on the Internet (such as news articles and this very blog post – both of which have a structure of their own).
[i] C. Fillmore, Frame Semantics and the Nature of Language, in: Annals of the New York Academy of Sciences, 280: 2032, 1976.
[ii] To give credit where it is due: this analogy was pointed out to me by Michael Ellsworth, a linguist at the International Computer Science Institute in Berkeley.
[iii] C. Levi-Strauss, The Structural Study of Myth, in: The Journal of American Folklore, Vol. 68, No. 270, Myth: A Symposium (Oct. – Dec., 1955), pp. 428-444, 1955
[iv] V. Propp, Morphology of the Folktale, Austin, University of Texas Press. Original Russian version 1928., 1968.
[v] Note that the or is exclusive.
[vi] Though Propp does note that these types introduce a helpful “Linnaean taxonomy” to the whole endeavor.
[vii] For example, Veselovskij suggests the motif “a dragon kidnaps the tsar’s daughter”; Propp points out that this can actually be abstracted into a general notion of some supernatural being (such as a devil or sorcerer) effecting the disappearance of a family member.
[viii] Propp enumerates 31 narrative functions in his survey of Russian folktales, not including the “initial situation”.
[ix] I think this is of particular interest, because it suggests a tendency of narrative elements to “cluster”. It’d be quite interesting to apply the tenets of construction grammar to a domain like narrative.