Aside from the probably most frequent use of the word which is to describe the feelings we undergo while watching a Hitchcock film for example, the word ‘suspense’ can also be used to describe the tension in a very close sports match, or even the rather nebulous waiting for an exam result. The fundamental metaphor at work suggests that something, perhaps a decision or an action, is ‘hanging in the air’ and has not yet ‘fallen to the ground’ or, in other words, become part of common knowledge. Dictionary definitions of the word ‘suspense’ suggest that the word designates a concept cluster than a single well defined concept. The Collins English dictionary  gives three definitions corresponding to different situations:
For this research, our goal is to create a model that can be used in all these suspenseful situations. We will exclude uses of the word which appear to be different to the meanings above, but will be looking for a more fundamental definition which reconnects the variety of definitions present in the literature.
A perusal of literature claiming to teach story structure and plot reveals surprisingly little mention of suspense. In ‘The Anatomy of Story’ , a major work by Truby on how to become a master story-teller, the word ‘suspense’ does not even occur once in 445 pages. Dibell’s book ‘Plot’  however, does mention suspense, and suggests three ways to produce it:
This relative dearth of information in the story-writing paradigm suggests that methods for producing suspense have remained at an intuitive level or that suspense is difficult to talk about in a precise manner.
The role of music in eliciting suspense in narrative is well-known. Certain film scenes would be much less interesting and suspenseful if there were no music accompanying them. As an example of this, we can mention a scene in Hitchcock’s film “Psycho” , where Janet Leigh is driving fast to escape the city, glancing regularly in her rear-view mirror. Seeing this scene with and without music produces very different effects. We believe that the music both triggers and signals a feeling of tension and anxiety. The spectator is affected by the mood of the music, but also interprets the presence of this music as a signal on the part of the film director that some highly significant and possibly dangerous event is imminent. This in turn heightens the attentiveness of the spectator to every small detail that might announce this event, and perhaps also makes the spectator imagine more vividly the different possibilities of what might happen.
Music can have the function of signalling i) either that something is about to happen (imminence), or ii) that something is not what it seems (missing information). It achieves these goals often by using its own musical suspense techniques. In Western culture, there are a certain number of musical ‘suspense signalling clichés’ which have been established, such as: diminished chords, tremolo passages, sudden variations in volume and so on. In music theory, these constructions are analysed as having high instability, or unpredictability, and this feature is perhaps important in their evocation of suspense.
Another feature which can sometimes intensify suspense is musical silence (see a recent analysis by Fink of music in Hitchcock films ). Silence may work by creating a type of musical ‘plot lull’, a kind of emptied acoustic space which (perhaps metaphorically) triggers the expectation that something will fill the space and that a sudden change may be imminent. Plot lulls are one of the features we observed during the exploratory phase of this research (see 4.1).
So, suspense in narrative can clearly depend also on extra-narrative features like music. However, the media-specific contributions to suspense that such features make will not however be our concern here.
Musical structure has often been compared to a narrative form (see just one example . The theoretical language of music theory even uses the term suspense in different situations (‘suspended cadence’, ‘suspended fourth’…) along with references to suspense over larger time scales when we feel the build-up of the music towards a culminating high-point where ‘something must happen’. When it finally arrives, the high-point often also contains an element of surprise, and this of course mirrors the behaviour of many story plots.
Unlike suspense in story, however, musical suspense does not usually evoke the emotion of fear. It seems that fear needs a more concrete link to the real world in order to be evoked. Within the framework of this research, it remains to be seen whether i) analogies from the musical world could give useful insights for our model of suspense in story, or ii) our model of suspense could be applied in a music theoretical setting.
In what has been called ‘the standard account of suspense’, the cognitive appraisal paradigm, suspense is conceptualised above all as an emotional experience that occurs over time (see work by Ortony et al.). Here, the before and after of the actual suspenseful experience are analysed into antecedents and effects. The end of suspense creates emotional states which range from satisfaction or disappointment for hoped-for positive events and relief or anguish for previously feared negative events (See also  ).
In general, however, scientific definitions use a variety of concepts:
Other concepts also often get a mention:
This rather broad collection of concepts of suspense reveals perhaps the somewhat confused state of current knowledge about suspense. We hope through our research to contribute to a theoretical clarification of the concept.
To show the path we intend to follow, we describe the relationship of our research to the preceding definitions as follows:
We will now examine the oft cited suspense paradox to investigate what light this can shed on current knowledge of the phenomenon of suspense.
We can resume the suspense (or repeater’s) paradox by the following question:
“If a necessary condition for suspense is not knowing the outcome of a narrative, how can a spectator who already knows the outcome feel suspense on repeated viewings?”
Smuts lists four different solutions to the suspense paradox ( p.1):
Smuts then goes on to discuss the weaknesses and strong points of each of these explanations. Within the framework of our research, we believe that only Smuts’ theory provides hypotheses that can be tested formally. This fact alone suggests that the theory is worthy of closer examination. Smuts’ desire-frustration theory of suspense claims that:
“when we know something that could help a character that we care about stay alive, and we are unable to relay the information, we feel suspense. Our desire to make use of the information is frustrated—that is, we want to help, but there is nothing we can do.” ( p.285)
In this view, powerlessness is a necessary condition for suspense and suspense is based on a series of frustrated desires. As we are mere onlookers of the unfolding drama, we have no means to intervene in the story. Our situation is similar to real world situations where there is nothing we can do to affect a result: waiting for an exam result perhaps. According to Smuts, as soon as we act to satisfy our desire and resolve the predicament then suspense disappears. Smuts aims with this model to solve the suspense paradox: we still want the victim in a narrative to escape in a given episode, event though we know from previous viewings that he does not.
During the presentation of our suspense model, we will discuss a possible solution to the suspense paradox which our particular model of suspense affords (see 4.3).
An oft quoted necessary condition for suspenseful drama is a lack of important information. As Hitchcock says, “The audience knows that a given piece of information is missing, but does not know what it is.” . This feature would, however, perhaps be better described as triggering curiosity, rather than suspense. Of course, as most suspenseful narratives also seem to use in one way or another the notion of missing information, it does indeed seem that the notions of curiosity and suspense are strongly linked. In this regard, White in 1938 had already claimed that suspense is ‘prolonged curiosity’ (see our previous reference 4). Hitchcock’s view of suspense as lack of information highlights once more the difficulty of distinguishing these concepts and in making clear how they work together in narrative.
Mieke Bal ( p.114) and Edward Branigan ( p.75) from the field of asthetics and literature theory, formalised in 1996 a typology of possible relationships between the reader and characters in narratives. To distinguish the different narrative structures, they imagine asking questions of both reader and characters and determine which of the latter would know the answers. Their analysis provides the first step in answering one of our additional research questions, that is: How can ‘character suspense’, that is the suspense that we interpret a character as feeling, be formally linked to ‘spectator suspense’? (see above 1.2). The 4 different cases they came up with can be resumed as follows:
We can see that at least three structurally different ways to produce suspense can be distinguished. Interestingly, the television series “Columbo”  is a kind of detective story which is nevertheless an example of the case 2 above. The spectator is ‘allowed’ to see who the murderer is right at the very start of the story, and so knows more than Columbo, the detective. The suspense of the series is therefore based on the spectator’s desire to find out not ‘who the murderer is’ but rather ‘how Columbo will detect the clues which might lead him to discover the murderer’. This narrative situation would provide an interesting test case for any suspense theory.
We now describe work on narrative in the story-modelling paradigm.
One might expect work in the story-modelling paradigm to provide insight into questions of suspense. Models of narrative such as TALESPIN, MINSTREL and MEXICA and others, have indeed created stories which claim to have a degree of suspense. However, none of these systems give an explicit formal analysis of how this suspense is created. The focus is rather on the global story-modelling task and on the automatic generation of new narratives. Some researchers  do attempt to evaluate the suspensefulness of their system’s outputs, but suspense is often seen as just one of a set of by-products of story generation which must be present for a story to be interesting.
It is our view that by focussing only on the global task, such systems may suffer from a degree of arbitrariness in the choice of their theoretical story modelling apparatus. We believe that the presence of more systematic and fundamental approaches to suspense (and to other aspects of what makes a story interesting such as curiosity and surprise), could help to create a common ground for the evaluation of story modelling systems.
One system which does attempt to use suspense explicitly in its story generation process is the work of Cheong & Young.
Cheong & Young  proposes a narrative generating system that attempts to create narratives specifically designed to evoke suspense in the reader. The system uses a data plan structure which models the goals and actions of the characters who belong to the given story-world. Similarly to our work, the focus is on the suspense created by the story structure itself. Following claims in Brewer and Lichtenstein  that affective states in the reader are provoked by changing the order of presentation of the events in a story world, they attempt to manipulate the level of suspense in a story. For this they use the idea that the suspense level readers feel depends on the number and type of solutions they can imagine in order to solve the problems facing their preferred character: “the reader’s suspense is heightened when undesirable outcomes are likely to happen over preferred outcomes” ( p.2).
More recently, related work by Bae & Young  has examined the relation and possible complementarity between suspense and surprise in relation to plot. This work uses the concept of ‘disparity of knowledge’ about the story between the reader and the characters. Whereas for suspense, the reader often knows more than the characters, in the case of surprise the reader knows less than (at least) one of the characters. The model distinguishes the plot the reader currently believes is a good description of the narrative from another (more accurate) sequence of events of which an important protagonist is aware. At a crucial moment, the narrative forces the reader to revise their reading of the story, creating surprise.
In the general approach we hope to develop, a disparity of knowledge between reader and characters can always be created, as the deductive and predictive processes of the reader and the characters are separate processes. In this first approach, however, we will be concentrating on the modelling of the reader’s inferential processes.
In the preceding chapter, we have given an overview of definitions of suspense from a wide range of fields and attempted to show how they are related. There is a clear lack of consensus on theoretical approaches to suspense in the literature. One of the goals of this research will be to contribute to a precise formal definition of suspense. Our general premise will be that
and our working hypothesis is that