Comparing Plot Structure and Character Agency in Mimetic vs. Diegetic Narratives

IGEL2022
#1

:speech_balloon: Carmen Tu

:classical_building: McMaster University

Title: Comparing Plot Structure and Character Agency in Mimetic vs. Diegetic Narratives

Abstract (long version below): Traditional plot models, such as the five-act model and the hero’s journey, do not describe plot differences between mimetic and diegetic narratives. We present a cross-modal analysis of plot structure based on the character’s emotional trajectory in a story to quantify similarities and differences among folk tales, novels, films, and video games. We observed that plot length can be similar across diegetic and mimetic narratives, but that the emotional progression of a protagonist (e.g., their agency) varies significantly. Our analysis demonstrates that a plot model based on protagonist psychology can detect both similarities and differences in plot structure across various media formats.


:movie_camera:


:newspaper: Long abstract

Introduction
The present study is a novel cross-modal analysis of plot structure, examining both short vs. long forms of narrative and diegetic vs. mimetic forms. The analytical method is based on two related approaches to plot structure from our lab that examine the character mediation of plot structure. The more descriptive approach is called the Resonator model, which models the plot of a story as a series of up-and-down “shifts” reflecting the changing psychological state of the protagonist throughout the story. A positive shift represents a change in the direction of positive emotion, whereas a negative shift represents a change toward negative emotion. A plot is comprised of a series of oscillations between positive and negative shifts. The more explanatory approach is called the Embodied Plot model (Tu & Brown, 2020), which proposes that the structure of a plot is predicated on the protagonist’s problem-solving dynamics throughout a story. The model describes a sequence of 14 plot components, eight of which are protagonist-driven cognitive processes that are part of the protagonist’s problem-solving cycle, including emotional appraisals, motivation, decision making, goal formulation, planning, action, and outcome evaluation.

Brown & Tu (2020) analyzed 288 Grimm Brothers’ fairy tales and the corpus of 182 Afanasyev Russian folk tales to validate the Resonator model. They were able to define a number of plot shapes for these stories based on the oscillatory patterns of shifts. This previous work only examined short folk tales, and so the current cross-media study of plot structure employed our analysis methods to investigate not only folk tales, but novels, feature films, and video games, with the aim of comparing short vs. long formats, and diegetic vs. mimetic formats. To our knowledge, no other empirical study has compared the plot structure and character psychology of narratives formats across such a diverse range.

Methods
We had one rater analyze 12 novels, a second rater analyze 45 feature films, and a third rater analyze 9 video games. We selected what we believed would be the most challenging narratives to analyze in order to test the applicability of our plot models. Each rater graphically mapped out the sequence of major psychological shifts – and therefore the plot arc – of a story based on the psychological story grammar outlined in the Embodied Plot model. After doing so, each rater created a corresponding Resonator timeline of the protagonist’s emotional trajectory. From the combined analyses, we obtained the following measures for each story: 1) the total number of shifts, which reflects a story’s length, and 2) the proportion of the total number of shifts that are protagonist-driven, rather than storyworld-driven, which is a reflection of the protagonist’s agency.

Results
Short vs. long diegetic narratives: Folk tales had a mean of 4.7 emotional shifts, with 59% of them being protagonist-driven. By contrast, novels had a mean of 14 shifts, with 39% of them being protagonist-driven. Hence, the number of shifts reflects the length of a story.

Short vs. long mimetic narratives: Feature films – which had a mean run time of 1.9 hours – had a mean of 11 shifts, with 44% of them being protagonist-driven. By contrast, video games – which had a mean playing time of 22.8 hours – had a mean of 45 shifts, with 46% of them being protagonist-driven. As with the diegetic narratives, the number of shifts reflected the length of the story, thereby demonstrating that our plot models are able to quantify basic differences in story length across media.

Diegetic vs. mimetic narratives: Novels (diegetic) and films (mimetic) had a similar number of shifts (14 and 11, respectively), but the protagonists in films had a slightly larger degree of agency (44% vs. 39%, respectively). For the comparison between novels (diegetic) and video games (mimetic), video games had a significantly greater number of shifts than novels (45 vs. 14, respectively), reflecting their greater length. Protagonists in video games also had more agency than those in novels (46% vs. 39%, respectively). Interestingly, protagonists in folk tales (diegetic) had the largest degree of agency of all of the media formats (59%).

Conclusions
This cross-media study shows that the Embodied Plot and Resonator models are able to reveal key features of plot structure across narrative media. First, the number of emotional shifts of the protagonist in a story is a marker of the length of a narrative format, spanning from folk tales to films to novels to video games. Next, we were able to see significant differences in protagonist agency across media, from novels to films to video games to folk tales. The protagonists of folk tales were a stand-out with regard to their high degree of agency. The current exploratory analysis is a first attempt to quantify differences in plot structure and protagonist agency across narrative media that vary in both length and diegetic/mimetic formats.

IGEL2022 conference schedule
#2

Hi @carmen_tu and thank you for your interesting presentation. I see a lot of potential for the method you presented here. I would like to know more, though, about the selection of your corpus. You mentioned yourself in the last slides that there is a big discrepancy in the number of folk tales versus novels, films and video games. How do you think that has impacted your results? Additionally, I would be interested to learn what kind of video games you took into account in your corpus, as I can imagine that it is quite different to determine emotional shifts in videogames versus the other three domains, as well as quite different to determine such shifts accross different types of games (e.g., particularly with respect to the notion of agency). Could you tell us a little more about how you comprised your corpora? Thank you!

2 Likes
#3

Hi Moniek, Thank you for the great question! The differences in the sizes between the corpora were mainly due to constraints on time and resources. However, I will explain our selection process for each corpus. Folktales have the greatest number of stories because our original study was about classifying plot shapes in folktales, specifically the folktales by the Grimm Brothers and Russian folktales published by Alexander Afanasyev. We report our results in the paper: Brown, S. and Tu, C. (2020). The shapes of stories: A “resonator” model of plot structure. Frontiers of Narrative Studies, 6(2), 1-29. However, we wanted to further validate our two plot models by using them to analyze more complex narrative formats, such as novels, films, and video games. For novels, we chose narratives with nonlinear storylines and unreliable narrators, such as Slaughterhouse-Five and Fight Club. Novels took the longest time for our coders to code, hence the low n. For films, we chose some that were folktale adaptations, some that were novel adaptations, and some that were originals. For video games, we chose games that were narrative and “character-driven” with a clear protagonist in the game narrative, such as The Witcher. Granted, video games were the most difficult to analyze due to the overlap of emotions between the game player and character.

In this presentation, we have pooled all of the different types of films into a single category, but in the future, we would like to do a direct comparison between folktales and film adaptations of folktales to see how length and agency compares and likewise with film adaptations of novels.

In terms of how this may have impacted the results, I believe that the results for novels may have been the most impacted. Nonlinear narratives with unreliable narrators are sometimes “slower”, for a lack of a better word. In Slaugherhouse-Five, Billy reacts to the events around him, and he often says “so it goes,” almost like a motto of resignation. In the film, we see Billy act a lot more. Some of these novels are known as psychological thrillers, such as Fight Club, where the protagonist has a passive persona buts emotional appraisals and attitudes. So you bring up a great point. The novels that we chose are likely to have had a significant impact on the results. If we had chosen more action-driven novels, then perhaps the agency scores would be closer to the ones we see with films and video games.

2 Likes
#4

Hello Carmen,
thank you for that exciting presentation! I am very curious about the part on fairy tales and agency, as I have come across different aspects of this topic. Specifically, there have been questions about the gender of protagonists: Ruth Bottigheimer even claims that the Grimm’s have altered the tales with each new edition, taking away direct speech from female characters. Anne Duggan also notices that in 19-century renditions female protagonists have particularly little agency, perhaps reflecting the societal norms at the time. Did you notice any differences between fairy tales with male and female protagonists, or was gender not part of your research design?

1 Like
#5

Hi Carmen, such an interesting presentation, thank you for sharing. I have a similar question to Moniek. I also am wondering what 9 video games you chose. I see that you responded that all of them were narrative, character-driven with a clear protagonist like the Witcher. If you don’t mind sharing (my email is hness1@gsu.edu if you prefer), I’m wondering about other differences among the 9 games. I ask because, for example, The Witcher 3 allows for a lot more player choice during the game (e.g., I can complete quests in whatever order I choose) and has some impact on the narrative (I think three possible endings?). Something like The Last of Us also has a lot of player choice, but the narrative always ends the same way. Any Telltale game does not allow the player to interact all that much compared to the other games, but the narrative does differ quite a bit depending on player choices. Telltale games almost play like an interactive film, so what I’m really wondering is if video games like those would look more similar to your results for films.

Another question I had was, with novels and video games with so many more emotional shifts, did you see repeating patterns? You showed that W-shape on the graph, and I’m wondering for stories with 40+ emotional shifts if the pattern or shape of the shifts repeats?

1 Like
#6

Very interesting work!

This recent book may be relevant: The Shapes of Stories (free to download until 19th July 2022)

1 Like
#7

There are also data from computational detection of story arcs that you could reuse: e.g. Hedonometer (in the supplementary materials of the article you will find a list of the works analyzed with the corresponding story shape)

And Roman Klinger has many publications and datasets related to emotions detection and characters’ roles: Selected Publications – Roman Klinger's Homepage

#8

Hello Carmen! I enjoyed your presentation and the discussion on it. Thank you! Have you considered how morals or a didactic slant to a story might affect levels of agency? This thought occurred to me specifically related to folk tales, since your examples came from Grimm, where one often finds a moral. My suspicion would be that for a moral to be communicated to readers, the protagonist needs to have a good deal of agency - it’s harder to judge someone who just has things happen to them and doesn’t act. The other media you look at also potentially have moral or didactic content, though maybe less obviously presented than folk tales. So, I wonder if the degree to which an author/creator pushes this kind of messaging through their story can affect levels of agency and the shape of the plot?