A domain-independent model of suspense in narrative

Doust, R.A. (2015). A domain-independent model of suspense in narrative. PhD thesis The Open University.

DOI: https://doi.org/10.21954/ou.ro.0000ab68


Many computational models of narrative have focussed on the structure of the narrative world. Such models have been implemented in a wide variety of systems, often linked to characters’ goals and plans, where the goal of creating suspenseful stories is baked into the structure of each system. There is no portable, independently motivated idea of what makes a suspenseful story.

Our approach is instead to take the phenomenon of suspense as the starting point. We extend an existing psychological model of narrative by Brewer and Lichtenstein (1982) which postulates suspense, curiosity and surprise as the fundamental elements of entertaining stories. We build a formal model of these phenomena using structures we call narrative threads.

Narrative threads are a formal description of a reader’s expectations about what might happen next in a given story. Our model uses a measure for the imminence of the predicted conflict between narrative threads to create a suspense profile for a given story. We also identify two types of suspense: conflict-based and revelatory suspense.

We tested the validity of our model by asking participants to give step- by-step self-reported suspense levels on reading online story variants. The results show that the normalised average scores of participants (N = 46) are in agreement with the values predicted by our model to a high level of statistical significance.

Our model’s interface with storyworld knowledge is compatible with recent developments in automatic harvesting of world knowledge in the form of event chains such as Chambers and Jurafsky (2008). This means that it is in principle scalable. By disentangling suspense from specific narrative content and planning strategies, we arrive at a domain-independent model that can be reused within different narrative generation systems. We see our work as a signpost to encourage the further development of narrative models based on what we see as its fundamental ingredients.

Viewing alternatives

Download history


Public Attention

Altmetrics from Altmetric

Number of Citations

Citations from Dimensions

Item Actions