Stand‑up comedy shows are far more carefully structured and timed than they may first appear.
Research based on performances of well-known comedians such as Sara Pascoe, shows that while a performance may feel spontaneous - full of ums, stumbles and moments that look like mistakes - stand-up shows are carefully constructed.
To discover the patterns beneath comedy shows, Dr Vanessa Pope, a postdoctoral researcher at King's College London, developed a computational methodology that visualises and describes timing structures in live performance.
The research, published in PNAS Nexus, draws on computer-based methods used to analyse musical timing and expressivity, including for heartbeats, a field pioneered by Professor Elaine Chew, a professor of digital music theranostics in the Research Department of Cardiovascular Imaging at King's.
The computational method, called Topology Analysis of Matching Sequences (TAMS), identifies parts of a performance that are repeated across different shows and maps how their timing shifts from night to night.
Between 2017 and 2018 during her PhD at Queen Mary University of London, she collected data by recording tour performances of established comedian and author Sara Pascoe and Pierre Novellie - a successful up-and-coming comedian then, who is now an award-winning performer.
Dr Pope applied TAMS to the recordings and found clear and recurring structural features across performances, despite differences in venue, audience and atmosphere.
For Pascoe, recorded towards the end of a mature touring show, an average of 40% of sequences within each performance matched exactly those from another show, compared to around 14% for Novellie, as he developed material for a new show.
Each comedian's show had a different structure. Pascoe's show had performance-specific material at the beginning of her set, while later sections relied on tightly rehearsed routines delivered with highly consistent timing. Over time, Novellie's show gathered material around jokes that were maintained, not only in content but in their relative position in performance time, with more flexible timing between matching sequences.
The analysis also revealed that elements which appear accidental, such as hesitations, filler sounds or apparent slips, could be embedded in material repeated across performances, suggesting they are deliberate parts of a comedian's delivery.
As a theatre director, Dr Pope is no stranger to live performance herself. She believes the results of this study highlight the complexity and skill involved in performance, and the importance of human connection that AI cannot replicate.
Performed speech is a feature of everyday life, from political speeches to work presentations, even the way people introduce themselves. Tracking how elements of speech such as rhythm, tone, emphasis, get repeated or adapted over time, you start to see how rich and flexible our communication really is."
Dr Vanessa Pope, a postdoctoral researcher at King's College London
Dr Pope said: "This research shows how we can use technology to highlight the craft behind live performance. These days, expressive forms of performance like comedy are often treated as something you could just generate with AI. But what this work really shows is that performance grows out of human connection - the back‑and‑forth, the timing, the interactions."
