A new study published in Science Advances reveals that the success of a story can be predicted by the number and intensity of narrative reversals, moments when the fortunes of characters dramatically shift. By analyzing 30,000 stories across various formats, researchers found that narratives with more frequent and pronounced changes in emotional tone tend to be more successful.
Stories have been a cornerstone of human culture for thousands of years, captivating audiences and imparting values across generations. Despite centuries of storytelling, the precise reasons why some stories resonate deeply while others fail remain elusive. Scholars have proposed many theories about what makes a story compelling, with concepts like plot progression, dramatic tension, and the balance between highs and lows often highlighted. Yet, past research has struggled to empirically verify these ideas or to predict the success of stories based on theoretical frameworks.
This study aimed to address that gap. With advances in computational linguistics, researchers saw an opportunity to quantify elements of storytelling, particularly focusing on narrative reversals. These are moments when a character’s fortune shifts drastically, such as a protagonist going from success to failure or from despair to triumph. By analyzing the frequency and magnitude of these shifts, the researchers hoped to uncover a measurable pattern that could explain why certain stories succeed.
“In addition to working as a quantitative researcher, I’m also a novelist, with an MFA from the Iowa Writers’ Workshop and my second novel coming out in 2025. I believe that new computational-linguistics tools allow us, for the first time, to build an empirically testable theory of narrative that can afford us new insights into why stories affect us the way they do,” said study author Samsun Knight, an assistant professor of marketing at the University of Toronto.
The researchers collected data from four distinct domains: movies, TV shows, novels, and fundraising pitches. Each dataset included a large sample of stories—3,713 movies, 19,339 TV episodes, 8,663 novels, and 1,133 GoFundMe campaigns. They used computational tools to track the emotional tone, or valence, of these narratives. Valence measures how positive or negative a text is, based on the words it uses. For example, words like “beautiful” and “love” score as positive, while terms like “doomed” and “disaster” score as negative.
To detect narrative reversals, the researchers broke down each story into fixed intervals of time or text. For movies, this meant dividing scripts into 5-minute windows, while TV shows used 3-minute windows, novels used 10,000-word windows, and fundraising pitches were analyzed with 500-word windows. The researchers then tracked how the emotional tone of the narrative changed over time. A narrative reversal was defined as a significant shift from a positive state to a negative state, or vice versa, within these windows. The number of these reversals and their magnitude (the extent of the emotional shift) were calculated for each story.
To ensure robustness, the researchers also tested the effect of other narrative features, such as sentiment volatility (how much the emotional tone fluctuated), word count (to control for the length of the narrative), and semantic properties like the complexity of language. Additionally, they controlled for factors specific to each domain, such as the budget of movies, the release year for TV shows, and the category of fundraising pitches. These controls helped the researchers isolate the effect of narrative reversals on story success.
Finally, the success of each narrative was measured differently depending on the domain. For movies and TV shows, success was gauged by ratings on IMDb, a popular movie and TV rating platform. For novels, success was measured by the number of downloads from Project Gutenberg. For fundraising pitches, success was defined by whether the pitch met its fundraising goal on GoFundMe.
Across all four types of narratives, the results were consistent: stories with more narrative reversals and more dramatic shifts in emotional tone were more successful. For movies, the number and magnitude of reversals predicted higher ratings on IMDb. TV shows followed a similar pattern, with episodes containing more dramatic reversals receiving higher audience ratings. Even in novels, where only text could be analyzed without the benefit of visual or auditory elements, stories with more reversals were downloaded more often from Project Gutenberg, indicating greater popularity.
For instance, movies with the highest number of narrative reversals saw an increase in their IMDb rating by up to 1.4 stars—an effect comparable to a $40 million increase in the movie’s budget. Similarly, novels with more reversals saw over a 100% increase in downloads, and fundraising campaigns experienced a nearly 40% higher chance of success when their stories had frequent and significant reversals.
The findings extended to amateur storytelling as well. In the context of fundraising pitches on GoFundMe, campaigns with more frequent and significant narrative reversals were more likely to meet their fundraising goals. This suggests that the power of narrative reversals isn’t limited to professional, carefully crafted stories but applies even in everyday storytelling contexts.
“Stories with more back-and-forth movement between good and bad fortune tend to be more engaging,” Knight told PsyPost. “For example, if you’re trying to fundraise for an illness, you might imagine choosing between two different ways of telling your story: 1. Describing how bad things are since your illness, and then talking about how badly you need support from friends and family, and how much worse things will get if you don’t get enough fundraising support; or 2. Describing how well things were going before, until the illness made everything turn for the worse, and then describing how much better things would be if you were able to fundraise successfully.”
“According to our findings, the story structure of (1) would likely be less effective, all else equal—a flatly negative story would not be as effective as a story with reversals, going from good to bad to good, as in (2).”
But as with all research, there are some caveats to consider. While computational tools like the VADER lexicon used in this study are effective at measuring valence, they cannot fully capture all nuances of human emotion. For example, a story may feature morally ambiguous situations where good things happen to bad people, which the computational analysis might struggle to interpret correctly. As such, the measurement of reversals may not always align perfectly with how a human audience perceives the narrative.
Future research could address these limitations by conducting controlled experiments where specific elements of a story are altered while others are kept constant. This would help to isolate the exact impact of narrative reversals on audience engagement. Additionally, researchers could explore other dimensions of narrative structure, such as pacing, the regularity of reversals, or how the size of reversals evolves over the course of a story.
Another promising avenue for future research is investigating the psychological mechanisms behind why narrative reversals are so effective. The study hypothesizes that reversals keep audiences engaged by creating tension and surprise, but more work is needed to confirm these assumptions. Understanding the emotional and cognitive processes that drive a story’s appeal could deepen our understanding of what makes narratives resonate with audiences on a fundamental level.
The study, “Narrative reversals and story success,” was authored by Samsun Knight, Matthew D. Rocklage, and Yakov Bart.
